HIL: Ctrl-C terminates the board worker before its lock task logs the failure #4500

A Ctrl-C on test/hil/hil_test.py leaves the board's lock record stale instead of ending its task as failed. The
run also writes no report.

Evidence

• ci.lan, raspberry_pi_pico, Take HIL board locks through hil_lock.board_task #4499's branch with agentrc 48d46bbf. SIGINT went to the run's process group after 4
  tests:
  • The run exited 130 with no report.
  • The lock record stayed stale (cleanupOwed: true, the board worker's pid), and no failed event was logged.
  • All 14 idle pool workers printed KeyboardInterrupt tracebacks. The board worker printed nothing.
• The next run on the board reported it as previous-cleanup-failure (found-left). Its park logged restored for
  it, so recovery works; only the interrupted run's own record of the failure is lost.

Cause

This is read from the source and is not yet confirmed by a trace:

• Main's KeyboardInterrupt in drain_pool is not caught by except Exception in main().
• Leaving with pool: therefore calls Pool.terminate(), which sends SIGTERM to every worker.
• The board worker has no SIGTERM handler, so it dies before its board_task can end the task and log the failure.
• _abort_report is skipped for the same reason, which is why there is no report.

Remaining work

• On KeyboardInterrupt in main(), let in-flight workers end their board task (bounded join, or close() +
  join(timeout)) before terminate(). Write the abort report as the other abort paths do.
• Confirm on hardware that a Ctrl-C then logs a failed event with the run's receipts and leaves no stale record.

Deferred: the pool shutdown predates #4499 and the recovery path covers it.