fix/client resync mechanism & server persist leader & quality-of-life improvements

[stanley2058]

This PR includes:

• Client would now try to resync whenever it detects an update is not being ack-ed from the server for over 3s
  • Side note: double-submitting is not an issue, yjs ignores duplicated updates.
• Server persistent leader
  • Use a Redlock-like mechanism to ensure persistDoc is only called once per namespace.
• Allow supplying a full socket instance to the client provider
  • This allows patching methods for testing; one MUST make sure the connection string is correct, since it's different from the one passed to the provider.
• Fix tsup broken due to incorrect Redis import
• Silent Redis client closed error (improve SNR in automated testing)