This is an example NVMe cleaner script that can listen to notifications
from a per-compute topic and perform basic NVMe (or generic command)
cleaning as a result.
It also supports periodic cleaning, notification pooling (if desired,
but not recommended), and single-shot (i.e. cron) mode.
I ran this while hammering my devstack node in a tight loop of create/
delete with a PCI device. The creates would succeed and fail as the
cleaning took a few seconds per iteration, then recover, create again,
delete, clean, etc. Dozens of loops in a row with expected results.
Related to blueprint one-time-use-devices
Change-Id: Ie4e81d6319499a3bf12b60e0af4cdb291fca9503
28331fee3c