Abort correct thread on TimeCheck timeout

The current TimeCheck implementation leads to incorrect
clustering/duping based on the abort stack trace being generated from
TimeCheck (running in the watchdog thread).

Appropriately signal the timed-out thread, so we receive more accurate
stack traces.

Test: atest timecheck_tests, manual verification of correct thread stack
trace.
Bug: 239252933
Bug: 227594853

Change-Id: I5d62581682ae85af269e7e10125343f096e26303