fix(iace): lower EstimateFrequency tiers — engine F was ~1 too high vs the GT

Diagnosis: engine F mean 3.56 vs professional 2.56; the dominant disagreement was
normal-operation hazards getting F=4 where the professional assigned 2. Lowered
the lifecycle→F mapping (normal operation 4→3, occasional phases 3→2). New
TestGT_RiskComparison_CrossGT runs the exact production comparison on BOTH GTs:
F within±1 rose to 95% (robot cell) and 94% (lift) — generic, not lift-tuned.

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

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

@@ -178,15 +178,17 @@ func EstimateFrequency(phases []string) int {
 		}
 		return false
 	}
+	// Calibrated to the professional's scale: the GT assigns lower exposure
+	// frequencies than a naive "operating = high" mapping. Normal operation is
+	// 3 (regular exposure), occasional phases (setup/maintenance/cleaning) 2,
+	// otherwise 2. (Engine F was systematically ~1 too high vs the GT.)
 	switch {
 	case has("normal_operation") || has("auto_operation") || has("manual_operation"):
-		return 4
-	case has("setup") || has("maintenance") || has("cleaning") || has("changeover"):
 		return 3
-	case len(phases) > 0:
+	case has("setup") || has("maintenance") || has("cleaning") || has("changeover"):
 		return 2
 	default:
-		return 3
+		return 2
 	}
 }