Fix test_from_worker_address_error[unreachable] test

The error raised may manifest in different ways depending on the
ordering and timing actions occur. An approach that seems to prevent
timing issues is to wait for the remote process to terminate entirely,
and not only depend on the `ucxx.reset()` call.

python/ucxx/ucxx/_lib_async/tests/test_from_worker_address_error.py

@@ -50,6 +50,9 @@ def _test_from_worker_address_error_client(q1, q2, error_type):
     async def run():
         # Receive worker address from server via multiprocessing.Queue
         remote_address = ucxx.get_ucx_address_from_buffer(q1.get())
+        if error_type == "unreachable":
+            server_closed = q1.get()
+            assert server_closed == "Server closed"
 
         if error_type == "unreachable":
             with pytest.raises(
@@ -177,6 +180,10 @@ def test_from_worker_address_error(error_type):
     )
     client.start()
 
+    if error_type == "unreachable":
+        server.join()
+        q1.put("Server closed")
+
     join_processes([client, server], timeout=30)
     terminate_process(server)
     try: