fix(validation): auto-cleanup stuck validation runs on backend startup

When the backend restarts (deploy, crash, reload), the in-memory
TaskManager loses its queue. ValidationRun records left 'running'
in the DB would hang forever, blocking re-runs.

Fix: during lifespan startup, query all ValidationRun with
status='running' and force-stop them as FAIL with a clear summary.

This prevents the 'already has a running run' error after restarts.

backend/src/app.py

@@ -84,6 +84,27 @@ async def lifespan(app: FastAPI):
         init_db()
         logger.info("👤 Bootstrapping admin user...")
         ensure_initial_admin_user()
+        # Clean up stuck validation runs from previous backend lifetime
+        # (runs left "running" in DB when the in-memory task queue was lost)
+        try:
+            from sqlalchemy.orm import Session as _Ses
+            from src.core.database import SessionLocal as _Db
+            from src.models.llm import ValidationRun as _VR
+            from datetime import datetime, timezone
+            _s: _Ses = _Db()
+            _stuck = _s.query(_VR).filter(_VR.status == "running").all()
+            for _r in _stuck:
+                _r.status = "FAIL"
+                _r.finished_at = datetime.now(timezone.utc)
+                _r.summary = "Force-stopped: backend restarted while run was in progress"
+                logger.reason(
+                    f"Force-stopped stuck run {_r.id} (policy={_r.policy_id})",
+                    extra={"src": "app.startup", "run_id": _r.id},
+                )
+            _s.commit()
+            _s.close()
+        except Exception as _e:
+            logger.warning(f"Failed to clean up stuck validation runs: {_e}")
         logger.info("⏰ Starting scheduler...")
         scheduler = get_scheduler_service()
     scheduler.start()