Handle and re-raise bunny exceptions during work

[shashankmehra]

Custom workers which perform bunny related operations inside the work function through the handler can find themselves getting stuck when rabbitmq is restarted.

This can happen because the bunny exception raised by the work function is caught by the catch all rescue present in process_work function - and this does not let the exception bubble up leaving bunny in broken state and not receiving any more jobs.

This PR allows Bunny::Exception to be re-raised so that the worker does not get stuck.

My use case involves allowing the ActiveJob to configure whether ack would be sent before or after the job (so that the Job can decide on at lease once or at most once paradigm). The jobs which were sending ack early through the work function would raise this exception and the worker would get stuck.

[michaelklishin]

I cannot reproduce the Ruby 3.2.x failure locally.