Serialize metadata cursor queries via ExplorerSchemaService and use in Tree mixin

[koianaoki]

Motivation

• Ensure raw cursor-returning metadata queries (e.g. MySQL SHOW statements) are executed through the session's serialized executor so they participate in the same retry/reconnect logic as other schema operations.
• Avoid direct, unsynchronized access to the raw connection from UI code and keep error/retry behavior consistent for explorer metadata operations.

Description

• Add ExplorerSchemaService.execute_cursor_query(query, database) which runs a cursor query through the service's serialized executor, returns (columns, rows), and always closes the cursor.
• Update TreeMixin._fetch_cursor_result signature to accept an optional database and prefer calling schema_service.execute_cursor_query(...) when available, falling back to the direct connection otherwise.
• Pass the selected node's data.database when invoking _fetch_cursor_result for MySQL SHOW FULL COLUMNS and SHOW INDEX code paths.
• Update tests to reflect serialized execution: modify tests/ui/explorer/test_table_metadata_shortcuts.py to assert calls to execute_cursor_query and add tests/unit/test_explorer_schema_service.py to verify executor usage and cursor closure.

Testing

• Updated UI tests in tests/ui/explorer/test_table_metadata_shortcuts.py were run and assert that execute_cursor_query is invoked for MySQL SHOW queries and that error paths notify appropriately, and they passed.
• New unit test tests/unit/test_explorer_schema_service.py was run and verifies that execute_cursor_query schedules work on the session executor and closes the cursor, and it passed.

---

Codex Task