matrix: judge-gated Shape B inject — closes lift-suite tail issues

Lift suite run #004 left two unresolved tail issues:
- Q6 ("Forklift loader") ↔ Q7 ("Hazmat warehouse, cold storage")
  swap recordings as warm top-1 because their embeddings are within
  0.20 cosine of each other. Distance gate can't tell them apart.
- Q9 + Q15 lose paraphrase recovery when qwen2.5 rephrases past the
  0.20 threshold. Distance says "drift too far"; sometimes the drift
  is real (skip), sometimes the paraphrase is still on-domain (don't
  want to skip).

Multi-coord run #008's judge re-rating proved the LLM can
distinguish: Q3 crane case landed at distance 0.23 (looks tight)
but rating 1 (irrelevant). The judge sees domain mismatch the
embedder doesn't.

This commit lifts that pattern into the matrix substrate. Shape B
inject now optionally routes every candidate through a judge gate
before the rank insert lands. Distance + judge BOTH have to approve.

internal/matrix/playbook.go:
- InjectPlaybookMisses signature gains a query string + an
  optional InjectGate. nil gate preserves pre-judge-gating
  behavior (current tests already pass with nil).
- New InjectGate interface + InjectGateFunc adapter for tests
  and non-LLM callers.
- Per-candidate gate.Approve(query, hit) call inserted between
  the dedup and the inject. Rejected candidates skip silently;
  injected count reflects post-gate decision.

internal/matrix/judge.go (new, ~140 lines):
- LLMJudgeGate calls an Ollama-shape /api/chat endpoint with the
  same 1-5 staffing-rubric prompt that worked in multi_coord
  run #008. fail-closed on HTTP/JSON errors (don't inject if
  judge can't speak — better miss than wrong-domain).
- NewLLMJudgeGate returns nil when URL or Model is empty,
  matching InjectGate's nil-means-no-judge semantics.

internal/matrix/retrieve.go:
- SearchRequest gains JudgeURL, JudgeModel, JudgeMinRating
  fields. Run() builds an LLMJudgeGate when set; passes nil
  otherwise. Backward compatible — existing callers see no
  behavior change.

Tests:
- TestInjectPlaybookMisses_GateRejectsCandidate (rejectAll → 0
  injected, even with tight distance)
- TestInjectPlaybookMisses_GateApprovesCandidate (approveAll →
  same as nil-gate behavior)
- TestInjectPlaybookMisses_GateSeesCorrectQuery (gate receives
  CURRENT query + RECORDED query separately so it can score
  the (current, candidate) pair)
- All 5 existing inject tests updated to new signature

go test ./internal/matrix → all 8 inject tests pass.
go test ./internal/matrix ./internal/shared ./cmd/{matrixd,
queryd,pathwayd,observerd} → all green.

STATE_OF_PLAY:
- OPEN item #1 (judge-gated injection) closed.
- DO NOT RELITIGATE adds the substrate-level judge-gate lock.
- OPEN list now 5 rows (was 6).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

internal/matrix/playbook.go

@@ -202,7 +202,24 @@ type PlaybookHit struct {
 // playbook-corpus cosine distance exceeds it are skipped (the boost
 // path may still re-rank them in place). Pass 0 (or any non-positive
 // value) to use DefaultPlaybookMaxInjectDistance.
-func InjectPlaybookMisses(results []Result, hits []PlaybookHit, maxInjectDist float32) ([]Result, int) {
+//
+// gate is an optional approval callback called once per CANDIDATE
+// (post-distance-filter, post-dedup) before injection. Returning
+// false rejects that candidate. Use nil for the historical "all
+// distance-eligible candidates inject" behavior.
+//
+// Multi-coord run #008's judge re-rating proved that distance + LLM
+// rating disagree often enough to matter (Q3 crane: dist 0.23 looks
+// confident, judge says 1/5 = irrelevant). Lift-suite tail issues
+// (Q6↔Q7 swap, Q9/Q15 paraphrase drift) are exactly this shape —
+// embedding-tight but wrong-domain. The gate parameter lets callers
+// route those candidates through a judge before the inject lands.
+//
+// query is the current search's query text — passed to the gate so
+// it can score (query, candidate) pairs without re-deriving from
+// SearchRequest. Empty when the caller doesn't have it (gate
+// implementations should treat empty query as "skip judge, allow").
+func InjectPlaybookMisses(query string, results []Result, hits []PlaybookHit, maxInjectDist float32, gate InjectGate) ([]Result, int) {
 	if len(hits) == 0 {
 		return results, 0
 	}
@@ -235,7 +252,16 @@ func InjectPlaybookMisses(results []Result, hits []PlaybookHit, maxInjectDist fl
 		}
 	}
 
+	injected := 0
 	for _, h := range bestForKey {
+		// Judge gate (per OPEN item #1, closed by this commit):
+		// post-distance-filter, ask the gate whether the candidate
+		// actually fits the current query before letting it inject.
+		// Closes the lift-suite tail issues where embedding said
+		// "tight" but a judge said "wrong domain."
+		if gate != nil && !gate.Approve(query, h) {
+			continue
+		}
 		injectedDist := h.Distance * float32(h.Entry.BoostFactor())
 		// Synthesize metadata that flags the injection so callers
 		// (driver/UI/observer) can distinguish "regular retrieval"
@@ -256,11 +282,34 @@ func InjectPlaybookMisses(results []Result, hits []PlaybookHit, maxInjectDist fl
 			Distance: injectedDist,
 			Metadata: meta,
 		})
+		injected++
 	}
 
-	return results, len(bestForKey)
+	return results, injected
 }
 
+// InjectGate is the optional approval callback for Shape B inject.
+// Called once per candidate (after distance filter, after dedup).
+// Returning false rejects that candidate. Implementations:
+//   - LLMJudgeGate (this package, see judge.go): Ollama LLM rates the
+//     (query, candidate) pair against a 1-5 rubric.
+//   - InjectGateFunc (this package): zero-deps adapter for arbitrary
+//     caller logic — useful in tests + when callers want non-LLM
+//     gating (e.g. metadata-only filters).
+//
+// nil InjectGate = pre-judge-gating behavior (all distance-eligible
+// candidates inject); preserves backward compatibility.
+type InjectGate interface {
+	Approve(query string, hit PlaybookHit) bool
+}
+
+// InjectGateFunc adapts a plain function to the InjectGate interface.
+// Used heavily in tests; production callers usually use LLMJudgeGate.
+type InjectGateFunc func(query string, hit PlaybookHit) bool
+
+// Approve makes InjectGateFunc satisfy InjectGate.
+func (f InjectGateFunc) Approve(q string, h PlaybookHit) bool { return f(q, h) }
+
 // ApplyPlaybookBoost re-ranks results in place using matched
 // playbook hits. For each hit whose (AnswerID, AnswerCorpus)
 // matches a result, multiply that result's distance by the hit's