Mock trading crashes after update_interval seconds: AsyncCache._sync_pnl was removed but MockLinearConnector still calls it

[bigsongeth]

Bug description

Any strategy using MockConnectorConfig (mock/paper trading) crashes update_interval seconds after start:

AttributeError: 'AsyncCache' object has no attribute '_sync_pnl'

The same error is raised a second time during shutdown: engine.dispose() → MockLinearConnector.disconnect() also calls self._cache._sync_pnl(...), so _dispose() aborts before cache.close() runs and the final orders/positions/balances sync to the storage backend is silently skipped.

Steps to reproduce

1. Run the bundled example: python strategy/binance/mock_trading.py (it configures MockConnectorConfig with update_interval=20).
2. Wait ~20 seconds. The periodic PnL task dies with the AttributeError above.
3. Stop the engine (Ctrl-C): engine.dispose() logs Dispose error: 'AsyncCache' object has no attribute '_sync_pnl' and skips the final cache sync.

Root cause

The storage-backend refactor (commit 070609b, "Add PostgreSQL support and update dependencies", issue #27) moved AsyncCache's direct SQLite access into pluggable StorageBackend classes and dropped AsyncCache._sync_pnl() in the process. But:

• MockLinearConnector still calls it in two places: _handle_pnl_update() (nexustrader/base/connector.py#L823) and disconnect() (#L832)
• Both SQLiteBackend and PostgreSQLBackend still create the {table_prefix}_pnl table — only the write path was lost, so the table stays empty forever.

Suggested fix

Restore the PnL write path through the backend abstraction (keeping the MockLinearConnector call sites unchanged):

• StorageBackend: add sync_pnl(timestamp, pnl, unrealized_pnl) to the abstract interface
• SQLiteBackend / PostgreSQLBackend: upsert the snapshot into the existing {table_prefix}_pnl table (INSERT OR REPLACE / ON CONFLICT (timestamp) DO UPDATE)
• AsyncCache: restore _sync_pnl() delegating to self._backend.sync_pnl(...)

I have this implemented with regression tests and can open a PR.