drt: add operation for slow disks

We currently have an operation which stalls disks on a given node until that node is killed by the disk stall detector. This certainly handles a class of issues that we want to stress, but it leaves another class uncovered - persistently _slow_ disks.

We should add an operation which dramatically slows down one or more disks on a given node (say, to 10% of its original speed). The hope is that this will uncover issues that are more commonly encountered by our customers (i.e. stalls which are short enough to cause a disruption, but not long enough to end up killing the node).

Jira issue: CRDB-43276