Capture disk usage metrics

[bloodearnest]

Now that we're on BindMountAPI in TPP, we can look at capturing detailed diskusage metrics or a job.

Previously, we had to use docker system df to track volume sizes, which was slow and added to the overloading the docker API problem.

But now we own the directories themselves, a simple du -d 1 /srv/high_privacy/volumes would gives us all volume sizes, which we can report on our spans just like CPU and memory (and record the high water mark).

Ideally, we'd track the total volume disk usage and also the new disk usage (which is current total - the disk usage post-PREPARE), as that might be a useful signal.