vine: largest task check does not consider used cache for disk

[btovar]

For some interval we check whether for every task there is worker large enough to run it. If not, then we print a warning. The check for disk only considers the total disk of a worker but not the used cache. This causes the warning not to be printed and for the workflow to make any progress.

[dthain]

As we discussed, there are a few potential situations that might lead to this:
1 - The system just does not have enough local disk capacity to run specific tasks.
2 - The workflow has filled up disks with intermediates that cannot be deleted. (workflow pruning will help here)
3 - The workflow has filled up disks with replicas of standard files that could be deleted.

Is there a straightforward way to address case 3?

For example, if we determine that a task cannot be scheduled for lack of disk space, can we search for a worker s.t. worker.sum(cached-normal-files) < task.input data size ?