Fix "coroutine was never awaited" warnings

[pentschev]

Fix warnings such as the one below:

sys:1: RuntimeWarning: coroutine 'BlockingMode._arm_worker' was never awaited
RuntimeWarning: Enable tracemalloc to get the object allocation traceback
Task was destroyed but it is pending!
task: <Task pending name='Task-22' coro=<BlockingMode._arm_worker() running at /opt/conda/envs/rn-230213/lib/python3.10/site-packages/ucp/continuous_ucx_progress.py:88>>

The fix requires introducing a fairly complex mechanism that checks for references to the UCP context and preemptively stop continuous progress when no listeners or endpoints are alive anymore. This is necessary because it's impossible to cancel a future once the event loop that owns it stops, which will happen in third-party projects where UCX-Py doesn't own the event loop, such as Distributed.

TODO: Verify how to still provide continuous progress for workers that do not make use of listeners, instead relying on ucp.recv(), which receives messages directly on the worker.