feat(ai-sdk): runnable iace-audit propose CLI + live LLM wiring (P2 slice 3)

Makes the offline proposer runnable end-to-end.

- BuildProposerInput (proposer_input.go): non-test engine->hazards path. The
  PatternMatch->Hazard converter is lifted out of the GT test files into
  production scope so both the tests and the CLI share one pipeline.
- iace-audit propose <narrative.json> [<ground-truth.json>]: detect candidates ->
  GT-screen survivors (when a ground truth is given) -> judge (HeuristicJudge by
  default, LLMJudge over ollama when IACE_PROPOSE_LLM=1) -> write the human-review
  queue to audit-reports/proposals.{md,json}. Propose-only.

Smoke run on a dishwasher narrative: 32 fired -> 3 candidates -> queue with a
confident duplicate, a confident distinct, and one punted to the LLM judge; GT
wall recall-safe. Live qwen is opt-in via env; the heuristic default keeps the
tool runnable (and CI deterministic) without a model. Proposal types 2-4
(foreign-framing gates, vocab->tag, coverage blind spots) remain for slice 4.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

ai-compliance-sdk/internal/iace/proposer_input_test.go

@@ -0,0 +1,25 @@
+package iace
+
+import "testing"
+
+func TestBuildProposerInput_WarewashingFires(t *testing.T) {
+	hazards, _, fired := BuildProposerInput(
+		warewashingNarrative,
+		"Gewerbliche Untertisch-Geschirrspuelmaschine (vernetzt)",
+		[]string{"food_processing"},
+	)
+	if len(fired) == 0 || len(hazards) == 0 {
+		t.Fatalf("want fired patterns + hazards, got %d patterns / %d hazards", len(fired), len(hazards))
+	}
+	has := func(id string) bool {
+		for _, pm := range fired {
+			if pm.PatternID == id {
+				return true
+			}
+		}
+		return false
+	}
+	if !has("HP2201") {
+		t.Errorf("warewashing-specific HP2201 must fire via BuildProposerInput")
+	}
+}