As per the mailing list discussion, this patch changes the watcher to
use a single job (two opcodes) for getting the cluster state (node list
and instance list); it will then compute the needed actions based on
this data.

The patch also archives this job and the verify-disks job.

Reviewed-by: imsnah