Fix the bbolt round trip, the blocklist gaps and the deploy set

The cache snapshot never loaded, so every restart was a cold cache. Snapshot
packed the whole key while Load decoded the record header out of the key's
remainder instead of the value, so every entry was skipped. The two sides are
now an explicit encodeRecord/decodeRecord pair with a version byte and a CRC-32C
trailer, so a future format change is detected instead of silently misparsed.
Load also drops entries that are already past ExpiresAt and rewrites the TTL to
the time actually left, so a served TTL never jumps back up across a restart.

Blocklist, all six gaps the failing tests were pointing at: a leading UTF-8 BOM
is stripped, "!allow example.com" works, SourceMeta exists for a source that
only ever contributed a rule, Pattern() reports the bare name again (the new
WireName() produces the wildcard form for the API), an RPZ record with no
explicit TTL inherits the zone default instead of whatever the reader left in
the RR, a hosts line with no address in the first field says so, and the
adblock parser handles CRLF and adblock-flavoured lines pasted into a plain
list. TestHolderSwapIsAtomic was asserting the wrong invariant — the swap was
always atomic, a name blocked by the old set is legitimately unblocked by the
new one.

Tests added for acl and resolver, both of which turned up real bugs:

- a PTR suffix matched across label boundaries, so a group scoped to
  example.com. also swallowed ample.com. and notexample.com. — a client could
  land in a group it does not belong to;
- the "*" catch-all was dropped from the rendered domain patterns, so the
  engine's per-query recompile lost the "everything" case;
- Upstream.Success never closed an open breaker, so a recovered upstream stayed
  excluded for the full cooldown;
- Exchange fanned out in parallel even under policy: failover.

Deploy set: go.mod requires Go 1.27.1 but CI pinned 1.23 and the Dockerfile
pinned golang:1.23-alpine, so neither could have built. The Dockerfile also
copied the source tree after the console bundle, which clobbered it.

The k8s set asked for hostNetwork with two replicas, but the server never sets
SO_REUSEPORT, so the second pod could not bind :53 and would crash-loop. It now
runs without hostNetwork behind a LoadBalancer Service, with a separate
ClusterIP Service for the Ingress (a headless Service cannot back one), a
ServiceAccount, an API-key Secret, per-pod PVCs, and a kustomization. The HPA
is gone: it cannot drive a StatefulSet. To let the API key come from that
Secret, config.Load now expands ${VAR} / ${VAR:-default} / ${NAME-default} and
fails to start on an unset variable with no default, so a missing secret is a
crash loop rather than an unauthenticated console. Expansion skips YAML
comments, so documenting ${VAR} in the ConfigMap does not break the file.

Verified: gofmt, go build, go vet clean; go test -race ./internal/... green
(acl, app, blocklist, cache, config, resolver, zonefile).

internal/cache/cache_test.go

@@ -1,6 +1,7 @@
 package cache
 
 import (
+	"encoding/binary"
 	"errors"
 	"math/rand/v2"
 	"net"
@@ -12,6 +13,7 @@ import (
 	"time"
 
 	"github.com/miekg/dns"
+	bolt "go.etcd.io/bbolt"
 )
 
 // fakeClock is a manually advanced clock so TTL, ageing and serve-stale
@@ -868,3 +870,231 @@ func BenchmarkManagerGetUnpack(b *testing.B) {
 		}
 	})
 }
+
+// TestStoreSurvivesProcessRestart is the test the warm-restart path exists for:
+// write a snapshot, close the file completely, open a brand new Store on the
+// same path and load it into a fresh LRU. Anything that only works because the
+// same *Store is still holding the file open fails here.
+func TestStoreSurvivesProcessRestart(t *testing.T) {
+	path := filepath.Join(t.TempDir(), "cache.db")
+	clk := newClock()
+	now := clk.Now()
+
+	st, err := OpenStore(StoreOptions{Path: path})
+	if err != nil {
+		t.Fatalf("OpenStore: %v", err)
+	}
+	warm := NewLRU(100, 2)
+	for i := range 5 {
+		n := "r" + strconv.Itoa(i) + ".example.com"
+		wire, _ := answer(t, n, 600, "10.1.0."+strconv.Itoa(i)).Pack()
+		warm.Put(key(n), &Entry{
+			Wire: wire, TTL: 600, Origin: "upstream", Upstream: "9.9.9.9:53",
+			CreatedAt: now, ExpiresAt: now.Add(600 * time.Second), StaleUntil: now.Add(630 * time.Second),
+			Bytes: len(wire),
+		})
+	}
+	if written, err := st.Snapshot(warm, now); err != nil || written != 5 {
+		t.Fatalf("Snapshot = %d, %v; want 5, nil", written, err)
+	}
+	if err := st.Close(); err != nil {
+		t.Fatalf("Close: %v", err)
+	}
+
+	// Second "process".
+	restarted, err := OpenStore(StoreOptions{Path: path})
+	if err != nil {
+		t.Fatalf("reopen: %v", err)
+	}
+	defer restarted.Close()
+	cold := NewLRU(100, 2)
+	loaded, skipped, err := restarted.Load(cold, clk.Now())
+	if err != nil {
+		t.Fatalf("Load: %v", err)
+	}
+	if loaded != 5 || skipped != 0 {
+		t.Fatalf("Load = %d loaded, %d skipped, want 5/0", loaded, skipped)
+	}
+	e, kind := cold.Get(key("r3.example.com"), clk.Now())
+	if e == nil || kind != Fresh {
+		t.Fatalf("restored entry: entry=%v kind=%v, want non-nil/Fresh", e, kind)
+	}
+	if e.Upstream != "9.9.9.9:53" {
+		t.Errorf("Upstream = %q, want 9.9.9.9:53", e.Upstream)
+	}
+	// The TTL a client sees must not jump back up across the restart: the
+	// loader rewrites it to the time actually left.
+	clk.Advance(120 * time.Second)
+	if got, want := e.AgeTTL(clk.Now()), uint32(480); got != want {
+		t.Errorf("AgeTTL after 120s = %d, want %d", got, want)
+	}
+}
+
+// TestStoreLoadSkipsDamagedRecords feeds Load a file where individual records
+// have been scribbled on, truncated or written by a future version. A snapshot
+// is a cache, not a source of truth: every one of these must cost exactly one
+// entry and none may fail the whole load.
+func TestStoreLoadSkipsDamagedRecords(t *testing.T) {
+	path := filepath.Join(t.TempDir(), "cache.db")
+	clk := newClock()
+	now := clk.Now()
+
+	st, err := OpenStore(StoreOptions{Path: path})
+	if err != nil {
+		t.Fatalf("OpenStore: %v", err)
+	}
+	l := NewLRU(100, 1)
+	for i := range 4 {
+		n := "d" + strconv.Itoa(i) + ".example.com"
+		wire, _ := answer(t, n, 600, "10.2.0."+strconv.Itoa(i)).Pack()
+		l.Put(key(n), &Entry{
+			Wire: wire, TTL: 600, CreatedAt: now, ExpiresAt: now.Add(600 * time.Second),
+			StaleUntil: now.Add(630 * time.Second), Bytes: len(wire),
+		})
+	}
+	if written, err := st.Snapshot(l, now); err != nil || written != 4 {
+		t.Fatalf("Snapshot = %d, %v; want 4, nil", written, err)
+	}
+	// Three different kinds of damage, applied to the first three records
+	// (the keys are packed, so the cursor is the only way to find them).
+	damaged := map[int][]byte{
+		0: futureVersionRecord,                  // written by a hypothetical v2
+		1: {recordVersion, 0, 0x01, 0x02, 0x03}, // truncated mid-varint
+		2: corruptRecord(t, encodeRecord(nil, &Entry{ // flipped bit
+			Wire: []byte{0xde, 0xad}, TTL: 1, CreatedAt: now, ExpiresAt: now, StaleUntil: now,
+		})),
+	}
+	if err := st.db.Update(func(tx *bolt.Tx) error {
+		b := tx.Bucket(bucketCache)
+		cur := b.Cursor()
+		k, _ := cur.First()
+		for i := 0; k != nil; i++ {
+			if bad, ok := damaged[i]; ok {
+				if err := b.Put(k, bad); err != nil {
+					return err
+				}
+			}
+			k, _ = cur.Next()
+		}
+		return nil
+	}); err != nil {
+		t.Fatalf("seed damaged records: %v", err)
+	}
+	if err := st.Close(); err != nil {
+		t.Fatalf("Close: %v", err)
+	}
+
+	again, err := OpenStore(StoreOptions{Path: path})
+	if err != nil {
+		t.Fatalf("reopen: %v", err)
+	}
+	defer again.Close()
+	restored := NewLRU(100, 1)
+	loaded, skipped, err := again.Load(restored, clk.Now())
+	if err != nil {
+		t.Fatalf("Load on a damaged file returned %v, want the good records to survive", err)
+	}
+	if loaded != 1 {
+		t.Errorf("loaded = %d, want 1 (only the untouched record)", loaded)
+	}
+	if skipped != 3 {
+		t.Errorf("skipped = %d, want 3", skipped)
+	}
+	if _, kind := restored.Get(key("d3.example.com"), clk.Now()); kind != Fresh {
+		t.Errorf("intact record did not survive: kind = %v", kind)
+	}
+	if _, kind := restored.Get(key("d0.example.com"), clk.Now()); kind != Miss {
+		t.Errorf("damaged record was loaded anyway: kind = %v", kind)
+	}
+}
+
+// futureVersionRecord is a well-formed record from a hypothetical v2 writer.
+var futureVersionRecord = func() []byte {
+	buf := []byte{recordVersion + 1, 0}
+	buf = binary.AppendUvarint(buf, 0)
+	return buf
+}()
+
+// corruptRecord flips one bit inside the body so the checksum no longer matches.
+func corruptRecord(t *testing.T, rec []byte) []byte {
+	t.Helper()
+	out := append([]byte(nil), rec...)
+	out[len(out)/2] ^= 0x40
+	return out
+}
+
+// TestRecordEncodeDecode pins the on-disk record format itself. The round trip
+// through bbolt used to break because the encoder and decoder disagreed; this
+// test fails at compile-review time rather than at warm-restart time.
+func TestRecordEncodeDecode(t *testing.T) {
+	now := newClock().Now()
+	in := &Entry{
+		Wire:       []byte{0x00, 0x01, 0x02, 0x03},
+		TTL:        600,
+		Origin:     "upstream",
+		Upstream:   "1.1.1.1:53",
+		CreatedAt:  now,
+		ExpiresAt:  now.Add(600 * time.Second),
+		StaleUntil: now.Add(630 * time.Second),
+		Neg:        true,
+		Truncated:  true,
+	}
+	out, err := decodeRecord(encodeRecord(nil, in))
+	if err != nil {
+		t.Fatalf("decodeRecord: %v", err)
+	}
+	if string(out.Wire) != string(in.Wire) {
+		t.Errorf("Wire = % x, want % x", out.Wire, in.Wire)
+	}
+	for _, f := range []struct {
+		name      string
+		got, want any
+	}{
+		{"TTL", out.TTL, in.TTL},
+		{"Origin", out.Origin, in.Origin},
+		{"Upstream", out.Upstream, in.Upstream},
+		{"Neg", out.Neg, in.Neg},
+		{"Truncated", out.Truncated, in.Truncated},
+	} {
+		if f.got != f.want {
+			t.Errorf("%s = %v, want %v", f.name, f.got, f.want)
+		}
+	}
+	// Times come back as wall-clock instants rebuilt from Unix nanoseconds, so
+	// they compare with Equal rather than == (the location bits differ).
+	for _, f := range []struct {
+		name      string
+		got, want time.Time
+	}{
+		{"CreatedAt", out.CreatedAt, in.CreatedAt},
+		{"ExpiresAt", out.ExpiresAt, in.ExpiresAt},
+		{"StaleUntil", out.StaleUntil, in.StaleUntil},
+	} {
+		if !f.got.Equal(f.want) {
+			t.Errorf("%s = %v, want %v", f.name, f.got, f.want)
+		}
+	}
+
+	// A record whose times are all zero must survive too: the varints encode
+	// them as 0 and come back as the zero time, not 1970.
+	zero, err := decodeRecord(encodeRecord(nil, &Entry{Wire: []byte{0xff}}))
+	if err != nil {
+		t.Fatalf("decodeRecord(zero times): %v", err)
+	}
+	if !zero.ExpiresAt.IsZero() || !zero.CreatedAt.IsZero() {
+		t.Errorf("zero times came back as %v/%v", zero.ExpiresAt, zero.CreatedAt)
+	}
+
+	for name, bad := range map[string][]byte{
+		"empty":          nil,
+		"header only":    {recordVersion, 0},
+		"no wire":        encodeRecord(nil, &Entry{}),
+		"bad checksum":   corruptRecord(t, encodeRecord(nil, in)),
+		"bad version":    futureVersionRecord,
+		"truncated tail": {recordVersion, 0, 0x80},
+	} {
+		if _, err := decodeRecord(bad); !errors.Is(err, errBadRecord) {
+			t.Errorf("decodeRecord(%s) error = %v, want errBadRecord", name, err)
+		}
+	}
+}