eliminate asyncio.sleep() and replace time.sleep() with a timeout

[alexrudd2]

asyncio sometimes behaves strangely on Windows, so longer delays were added to prevent flaky tests. However, time.sleep(10) will always wait the full 10s, even though we really only need to wait until the server is stopped with async_stop().

Using future.result with a timeout waits only as long as necessary. This speeds up the CI on Windows significantly.

Also, having an await asyncio.sleep() immediately after an await cls.active_server.server.shutdown() is redundant. Control flow will already yield back to the event loop with the first await

See #2033 (comment)

[alexrudd2]

We have had severe problems in the past, but not forcing a task switch which is essentially what asyncio.sleep(0) does. Not having a task switch mean e.g. that the task is cancelled but not yet terminated, not always but sometimes (especially on windows).

Originally posted by @janiversen in #2033 (comment)

I believe you were talking about the sync code. asyncio.run_coroutine_threadsafe(cls.async_stop(), cls.active_server.loop) will only schedule async.stop() and not wait until termination. However, this should be avoided by the use of future.result. If not, the "health check" will catch any regressions:

pymodbus/test/conftest.py

Lines 173 to 174 in a152762

	@pytest.fixture(name="system_health_check", autouse=True)
	async def _check_system_health():

[janiversen]

Looks like you are correct on that part.