[SPARK-55448][CONNECT] Fix query events loss when session closes during query execution

[biruktesf-db]

What changes were proposed in this pull request?

Allow Spark Connect to send query events after the session is closed.

Why are the changes needed?

A race condition exists between Spark Connect session closure and execution cleanup. When a session closes while an execution is still running:

1. SessionHolder.close() interrupts active executions via removeAllExecutionsForSession(), then transitions the session status to Closed via eventManager.postClosed()
2. The interrupted execution thread's finally block calls ExecuteHolder.cleanup(), which attempts to post terminal events (postCanceled, postClosed) through ExecuteEventsManager
3. ExecuteEventsManager.assertStatus() requires sessionStatus == SessionStatus.Started, but the session is now Closed by the original thread handling SessionHolder.close().
4. An IllegalStateException is thrown, silently swallowing the terminal events.
5. Any downstream listeners never receive the Canceled/Closed events for the operation, leaving it appearing as perpetually running

Does this PR introduce any user-facing change?

No.

How was this patch tested?

Added unit tests.

Was this patch authored or co-authored using generative AI tooling?

Generated-by: Claude Code v2.1.37