Harden ROLLBACK against I/O errors in transactions

If the original error (e.g., disk full) also prevents ROLLBACK from
executing, catch and suppress the ROLLBACK failure so the original
exception propagates cleanly and the connection isn't left in a
permanently broken state.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

db.py

@@ -295,7 +295,10 @@ class ProjectDB:
             self.conn.execute("COMMIT")
             return df_id
         except Exception:
+            try:
                 self.conn.execute("ROLLBACK")
+            except Exception:
+                pass
             raise
 
     # ------------------------------------------------------------------

utils.py

@@ -225,7 +225,10 @@ def sync_to_db(db, project_name: str, file_path: Path, data: dict) -> None:
 
             db.conn.execute("COMMIT")
         except Exception:
+            try:
                 db.conn.execute("ROLLBACK")
+            except Exception:
+                pass
             raise
     except Exception as e:
         logger.warning(f"sync_to_db failed: {e}")