Fix dispose_orm() not disposing async engine on shutdown

[kaxil]

dispose_orm() now disposes async_engine and clears AsyncSession, preventing connection leaks on process exit, gunicorn worker restarts, and atexit

Background

When _configure_async_session() was extracted from configure_orm() the corresponding cleanup in dispose_orm() was not updated. The async engine's connection pool was never closed on:

• atexit.register(dispose_orm) -- process exit
• Gunicorn worker_exit hook -- worker restarts
• reconfigure_orm() -- ORM reconfiguration

With the async engine now actively used by the Execution API (via get_team_name_dep in the dependency chain for connection/variable lookups), each worker restart leaked pool connections against PostgreSQL's max_connections.

Approach

Calls async_engine.sync_engine.dispose() (synchronous path) rather than await async_engine.dispose() since dispose_orm() runs in sync context. This matches the existing pattern in clean_in_fork().