[orchagent]: _exit() on graceful shutdown to avoid destructor-chain memory corruption

[Copilot]

On SIGTERM/SIGINT, OrchDaemon::start() returns and main() falls through to return 0, running the full ~OrchDaemon destructor chain. This is unsafe: FlexCounterManager destruction makes blocking SAI calls through sairedis's ZMQ channel while libzmq I/O threads are still alive, and orchs torn down earlier in the reverse-order loop have already freed buffers those threads still reference — corrupting the heap and causing intermittent jemalloc crashes.

Changes

• orchagent/orchdaemon.cpp — adds exit_if_graceful_shutdown_requested(): drains the async swss recorder (setAsync(false) joins the worker and flushes pending records), then calls _exit(0) to bypass the destructor chain entirely
• orchagent/orchdaemon.h — declares the function with _exit as the default; accepts an injectable exit_fn for unit tests
• orchagent/main.cpp — calls exit_if_graceful_shutdown_requested() immediately after orchDaemon->start() returns on the graceful path
• tests/mock_tests/orchdaemon_ut.cpp — unit tests covering the no-shutdown no-op and the drain+exit path

// After orchDaemon->start() returns on SIGTERM/SIGINT:
exit_if_graceful_shutdown_requested();  // drains recorder, _exit(0) — skips ~OrchDaemon

return 0;  // only reached on non-graceful exit

[mssonicbld]

/azp run

[azure-pipelines]

Azure Pipelines:
There may be pipelines that require an authorized user to comment /azp run to run.

[azure-pipelines]

Azure Pipelines:
Successfully started running 1 pipeline(s).