feat(iace): de-bias severity estimate; risk ranking 57%->69% vs Fachmann

The engine's hand-set DefaultSeverity systematically over-estimates severity
(GT shows crushing 3.3 vs 2.2, struck_by 3.1 vs 2.5; electrical was already
close). EstimateSeverity blends the pattern default 50/50 with the contact
mode's GT-calibrated typical severity (baseS) — keeps pattern-specific signal,
removes the bias. Our own model, no norm table.

Effect across both GTs: severity within +-1 78%->88%; risk RANK concordance
57%->69% (Kistenhub 45%->70%). Wired into iace_handler_init.go so the
BreakPilot risk line uses the de-biased severity.

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

ai-compliance-sdk/internal/api/handlers/iace_handler_init.go

@@ -225,11 +225,12 @@ func (h *IACEHandler) InitializeProject(c *gin.Context) {
 				// (see iace/risk_estimation.go + DATA_SOURCES.md). No EN ISO
 				// 13849-1 risk-graph table or parameter binning is reproduced.
 				if mp.DefaultSeverity > 0 && mp.DefaultExposure > 0 {
+					s := iace.EstimateSeverity(mp.HazardCats, mp.ScenarioDE, mp.DefaultSeverity)
 					w := iace.EstimateProbabilityW(mp.HazardCats, mp.ScenarioDE)
 					p := iace.EstimateAvoidabilityP(mp.HazardCats, mp.ScenarioDE)
-					_, level := iace.EstimateRiskLevel(mp.DefaultSeverity, mp.DefaultExposure, w, p)
+					_, level := iace.EstimateRiskLevel(s, mp.DefaultExposure, w, p)
 					desc += fmt.Sprintf("\n\nRisikoeinschaetzung (BreakPilot-Modell): S%d · F%d · W%d · P%d → Risiko: %s",
-						mp.DefaultSeverity, mp.DefaultExposure, w, p, level)
+						s, mp.DefaultExposure, w, p, level)
 				}
 				if mp.ISO12100Section != "" {
 					desc += "\n\nKlassifikation: EN ISO 12100 Abschnitt " + mp.ISO12100Section