fix(ws): per-instance event loop for same-process multi-Client

[liangxp1990]

ws/client.py used a module-level loop singleton (loop = asyncio.get_event_loop() at import). Client.start() did loop.run_until_complete(_select()) which occupied that single loop indefinitely, so a second Client in the same process collided with "This event loop is already running" — making same-process multi-bot impossible (e.g. agno AgentOS multi-bot pattern, OpenClaw-style multi-account).

Fix: each Client gets its own loop (self._loop = asyncio.new_event_loop() in init) and uses self._loop everywhere (start/_connect/_receive_message_loop/ _schedule_handle_message). Single-Client behavior unchanged; N Clients now coexist in one process, each on its own loop — aligning with OpenClaw (Node) which runs multiple accounts concurrently on one event loop natively.

Verified: two FeishuChannel (two app_id) in one agno AgentOS process both connect (running:true), no "event loop already running".

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}