Right now, the actor_states_monitor loop only works on PyActorMesh, and the handling / propagation of unhandled supervision events only works with PythonActor. We want to extend this to work for all Actor types, including rust actors, and not rely on python specifics.
This should help the robustness of error detection, and centralize how it is handled into a single implementation.
This will also make it easier to implement a push-based error handling mechanism rather than continuous polling