fix sorted ranking so model uses overall sorted retrieved chunks

src/rag/retrievers/schema.py

@@ -393,7 +393,7 @@ class SchemaRetriever(BaseRetriever):
         db_ranked = self._rank_db_tables(db_tbl_results, db_col_results, fts_results, top_k=k)
         tabular_ranked = self._rank_tabular_sheets(sheet_results, tabular_results, top_k=k)
 
-        results = db_ranked + tabular_ranked
+        results = sorted(db_ranked + tabular_ranked, key=lambda r: r.score, reverse=True)
         logger.info(
             "schema retrieval",
             count=len(results),