Async Postgres and MySQL deletes raise when they fail

Agno's AsyncPostgresDb and AsyncMySQLDb now log a failed delete_run and re-raise the error, the way their sync versions and the async SQLite and MongoDB backends already did. A dropped connection, a deadlock or a timeout used to come back as False, which callers read as "no such run" while the run stayed in the table. delete_runs logged the error and returned as if it had worked, and now raises too.

We ran both releases against an embedded Postgres, deleting a missing run as the owner and then deleting a stored run as a role that can read the runs table but can't delete from it:

Before: missing run   -> False
        failed delete -> False                         run-1 still in table: True
Now:    missing run   -> False
        failed delete -> raises ProgrammingError (psycopg.errors.InsufficientPrivilege)
                         permission denied for table agno_runs
                                                       run-1 still in table: True

Code that calls these methods directly can now catch the exception to retry or report the failure, and keep reading False as a run that was already gone.

See the cookbook, and learn more about async Postgres and async MySQL in the documentation.