Also removed unnecessary declarations of instance variables on
ProxManager, only keeping the ones that aren't set directly in
__init__ around for greater visibility.
noout will now be set on ceph before rebooting a node. Once the node
is back up, the script will wait until ceph reports no issues apart
from noout being set before unsetting the flag and moving on to the
next node.
The script will now consistently wait long enough after triggering a
reboot to ensure the node has gone down before starting to poll for
the node being back up again.
