#4 Compute client disk tidy thread

[Keith Flanagan]

We need a thread running within the compute client that starts to delete resources if the free disk space drops below a specified threshold level.

Only "unused" resources should be deleted. I think we can "renew" resources when they are "obtained" with the local resource cache via File.setModifiedTime() or somesuch so that frequently used resources don't get deleted.

What happens if this thread deletes published results of previous job executions before they have a chance to be backed up by the archiver? I think we'll just rely on the task/job scheduler to reschedule jobs to run somewhere again / somewhere else. I think we can assume that nodes should have sufficient disk space to perform the work they are required to.