#14 Implement pull subscriber for compute nodes

[Keith Flanagan]

At present, compute nodes can request that archiver nodes seed/reseed a particular torrent resource via a "please seed resource X"-type message.

However, compute nodes currently can't receive such messages because they aren't web services. We should implement a thread that periodically polls the notification system for "please seed torrent" messages so that compute nodes can respond to such requests.

Implementing this feature should increase the chances of successful job execution result archival in the cases where compute nodes have a high turnover of jobs. In these cases, although the results of computation are more than likely still available on the node's disk, the seeding torrent may have fallen off the list of active torrents, and so appear unavailable to the outside world.