feat: robust metric extraction with confidence score and proof snippets

- fixed Year-Prefix Bug in MetricParser
- added metric_confidence and metric_proof_text to database
- added Entity-Check and Annual-Priority to LLM prompt
- improved UI: added confidence traffic light and mouse-over proof tooltip
- restored missing API endpoints (create, bulk, wiki-override)

company-explorer/diagnose_wolfra.py

@@ -0,0 +1,41 @@
+import sys
+import os
+import logging
+
+# Add backend to path
+sys.path.append(os.path.join(os.getcwd(), 'backend'))
+
+from backend.database import SessionLocal, Company, Industry
+from backend.services.classification import ClassificationService
+
+# Setup basic logging
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger(__name__)
+
+def run_wolfra_reevaluation():
+    db = SessionLocal()
+    try:
+        # Find Wolfra
+        company = db.query(Company).filter(Company.name.ilike("%Wolfra%")).first()
+        if not company:
+            logger.error("Wolfra not found in DB")
+            return
+        
+        industry = db.query(Industry).filter(Industry.name == company.industry_ai).first()
+        if not industry:
+            logger.error(f"Industry '{company.industry_ai}' for Wolfra not found.")
+            return
+
+        logger.info(f"Starting targeted re-evaluation for {company.name} (ID: {company.id})")
+        
+        classifier = ClassificationService()
+        updated_company = classifier.reevaluate_wikipedia_metric(company, db, industry)
+        
+        logger.info("--- Re-evaluation Result ---")
+        print(f"Final Calculated Metric: {updated_company.calculated_metric_value} {updated_company.calculated_metric_unit}")
+
+    finally:
+        db.close()
+
+if __name__ == "__main__":
+    run_wolfra_reevaluation()