Add interrupted policy to MessageChannelPartitionHandler polling

[brian-mcnamara]

While testing shutdown handling, I found MessageChannelPartitionHandler would cause issues as the poller indefinitely poll until the steps complete. If the manager happens to be shutting down, this causes the job to become stuck.

This solution is a shim to better support graceful shutdown; I don't believe its a proper fix since, in my opinion, MessageChannelPartitionHandler should be refactored into something like TaskletStep which can manage repository updates during an execution lifecycle, however that requires a larger design / discussion.

For our use case, I plan on overriding the policy to throw JobInterruptedException when the pod is being shutdown.