fix QSM RestartNode race: wait for reconnect; remove debug logging
Add waitForNextConnection to the QSM RestartNode semantics so the test blocks until db-sync has actually reconnected to the restarted server before proceeding. This mirrors the fix already applied to the Babbage and Conway unit tests and closes the same race condition (the ~29s mux timeout + retry window on macOS CI). Also remove all temporary debug logging (putStrLn / traceM / Debug.Trace) from Server.hs that was added during diagnosis.