fix: Close DuckDB cursor to release internal buffers

Root cause: DuckDB cursor objects were never being explicitly closed,
causing internal buffers to accumulate in memory.

The pattern was:
```python
rows = conn.execute(sql).fetchall()
```

This creates a cursor object from `conn.execute(sql)`, fetches all rows,
but never closes the cursor. Even after `fetchall()` and `del rows`,
the cursor object holds references to DuckDB's internal result buffers.

These buffers accumulate across queries, causing steady memory growth
that GC cannot collect (they're managed by DuckDB's C++ layer, not Python).

Solution: Explicitly close cursor after fetching results

```python
cursor = conn.execute(sql)
rows = cursor.fetchall()
# ... process data ...
cursor.close()  # Release DuckDB internal buffers
del cursor
```

This ensures DuckDB's C++ result buffers are freed immediately after
the query data is extracted, rather than accumulating until the
connection is closed/reset.

Expected impact: Memory should stabilize immediately. Each query will
fully clean up after itself, preventing the steady growth pattern.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

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

Author: Ignacio Van Droogenbroeck

api/duckdb_engine.py

@@ -373,16 +373,21 @@ def _execute_with_pool(self, sql: str) -> Dict[str, Any]:
             with self.connection_pool.get_connection(timeout=5.0) as conn:
                 # Execute query
                 query_start = time.time()
-                rows = conn.execute(sql).fetchall()
-                columns = [desc[0] for desc in conn.description] if conn.description else []
+
+                # CRITICAL: Keep reference to cursor to close it properly
+                cursor = conn.execute(sql)
+                rows = cursor.fetchall()
+                columns = [desc[0] for desc in cursor.description] if cursor.description else []
                 query_time = time.time() - query_start
 
                 # Convert rows to list immediately
                 data = [list(row) for row in rows]
                 row_count = len(data)
 
-                # CRITICAL: Delete rows reference immediately to free DuckDB result memory
+                # CRITICAL: Delete ALL references to free DuckDB memory
                 del rows
+                cursor.close()  # Close cursor to release DuckDB internal buffers
+                del cursor
 
                 total_time = time.time() - start_time