Let each HarvestControllerServer manage its own Umbra instance

It is perfectly possible to have multiple independent Umbra processes sharing the same RabbitMQ broker on the same or different machine. It requires only that each umbra uses its own independent exchange/queues on the broker, and this can be configured on the command line when umbra is started.

Therefore it would be quite straightforward to have each HCS application manage its own independent Umbra instance - restarting it before and after each harvest job and flushing any items left over in the queues.

This would require some modification to the bean configuration in NAS to inject the right queuing information, and also the addition of two hooks - external scripts to be run before and after each harvest job.