fix: unref BrowserController.close() fallback timer to avoid keeping Node alive longer than needed

[simshaun]

Summary

This PR fixes an event-loop retention issue in browser-pool where BrowserController.close() schedules a 5 second fallback kill timer that can keep Node alive after crawl work has already finished.

The timer is created and left referenced. This change keeps the fallback behavior intact but calls unref() on that timer so it does not block process exit.

Why this is needed

In current behavior, a minimal crawl can complete and all user work can be done, but process exit is delayed due to the fallback timer.

Minimal reproduction

Run any short crawler that exits immediately after crawler.run() and cleanup. Observe that process exit occurs ~5 seconds later than expected.

[barjin]

Thank you for spotting this @simshaun! The fix looks reasonable to me 👍

I'm not really convinced about the test, as it's imo checking this specific implementation. Perhaps we don't need the test for this?

[janbuchar]

Thank you for spotting this @simshaun! The fix looks reasonable to me 👍

I'm not really convinced about the test, as it's imo checking this specific implementation. Perhaps we don't need the test for this?

Agree on this one, please get rid of the test