Fix Python process-pool test server lifecycle

[kpumuk]

Python process-pool tests currently stop the server through a four-second alarm and os._exit(0). The abrupt exit bypasses normal multiprocessing cleanup, can report leaked semaphore objects, and forces every process-pool test to wait for the alarm.

Observed macOS failures from this lifecycle include:

• macOS 15 Intel with Python 3.14, where the alarm shut down TProcessPoolServer while the client was still running and emitted leaked-semaphore warnings.
• macOS 15 Intel with Python 3.12, where the server exited with return code 0 before the harness completed.
• macOS 15 with Python 3.13, where the earlier process-pool shutdown path deadlocked until GitHub Actions cancelled it at the six-hour limit.

This change replaces the alarm with a SIGTERM handler that unwinds TProcessPoolServer.serve() through its existing SystemExit handling. The client/server harness waits for the complete POSIX process group to exit and falls back to SIGKILL only when graceful shutdown times out. Windows retains equivalent terminate-and-kill fallback behavior for its supported server types.

This is a focused test-harness lifecycle fix, so Jira was intentionally skipped.

• Did you create an Apache Jira ticket? (Request account here, not required for trivial changes)
• If a ticket exists: Does your pull request title follow the pattern "THRIFT-NNNN: describe my issue"?
• Did you squash your changes to a single commit? (not required, but preferred)
• Did you do your best to avoid breaking changes? If one was needed, did you label the Jira ticket with "Breaking-Change"?
• If your change does not involve any code, include [skip ci] anywhere in the commit message to free up build resources.