fix: gracefully handle redis counters via heartbeat

[jowparks]

I believe the issue we were seeing with hitting limits is that the websocket-proxy was getting interrupted for reboots and because the connections weren't gracefully terminated, the counter was not decremented. This handles the connection management via heartbeats and periodic cleanup by existing clients.

It adds functionality to reset Redis counters on boot, preventing stale counters from affecting new instances. The core improvement is a heartbeat system using UUID-based instance IDs with background tasks that periodically update instance status and clean up counters from stale instances. These changes provide more reliable rate limiting in distributed environments, automatically recover from instance crashes without manual intervention, and maintain accurate global and per-IP connection counts across multiple proxy instances.

[cody-wang-cb]

Some comments, but this approach makes sense to me overall