Most of them were not testing what they are supposed to test. The
exposed errors, for example, were due to a misconfigured Validate
mock, not because of actual deployment failures. Clean it up.

Hosts can stay in the "available" state for a long time. If the
underlying ramdisk image changes, their state may become invalid.

This change allows an operator to request a host reboot that will also
act in states where the ramdisk is waiting for something.

Allows restarting a running inspection on image change.

The "force" argument to InspectHardware has been renamed to
restartOnFailure to avoid confusion.

Change the error handling code to immediately retry aborted inspection.

The only way to do it is to go through cleaning.

The word "force" is too generic, the new name reflects better what
the flag does: restart the process when it previously failed.

Power actions are left intact since "force" has a more subtle dual
meaning in PowerOff.