Right now if you try to failover an instance which is not marked as up the
operation will fail unless you pass the --ignore-consistency flag because the
disks won't be considered to be consistent. Allow them to be if we know the
instance shouldn't be up.

Reviewed-by: imsnah