ensure correct lock type propagated in pool recreate

ensure correct lock type propagated in pool recreate

Fixed critical issue in asyncio version of the connection pool where
calling :meth:`_asyncio.AsyncEngine.dispose` would produce a new connection
pool that did not fully re-establish the use of asyncio-compatible mutexes,
leading to the use of a plain ``threading.Lock()`` which would then cause
deadlocks in an asyncio context when using concurrency features like
``asyncio.gather()``.

Fixes: #10813
Change-Id: I95ec698b6a1ba79555aa0b28e6bce65fedf3b1fe
(cherry picked from commit 2ed32bbf891b8f7e6c151071b4711319d9aa84f0)