During deprovisioning of a Host, if 'deleting' (i.e. deprovisioning) the
node succeeds (i.e. it doesn't go to the Error state) but the automated
cleaning that follows fails, the only way to recover is to return the
node to the manageable state.

Previously, once in the manageable state we would attempt adoption on
the node so that we could deprovision again. However, in the course of
'deleting' the node, the image information is cleared from it so it
cannot be adopted again. (Adoption continues to be the right thing to do
if the node has just been re-registered due to the Ironic database being
recreated, and in that case the image information is present since it
gets added during the initial registration.)

To work around this, don't attempt to adopt during the Deprovisioning
state if the node is manageable and the image data is not present.
Handle the manageable state in Deprovision() by declaring the
deprovisioning complete.

A node in the manageable state cannot be re-provisioned without first
being cleaned - it must go through cleaning to reach the available state
before it can be provisioned. Provisioning already handles nodes in the
manageable state, as this is how they begin after the initial inspection
of the host before the first provisioning (which does the initial
cleaning).

(cherry picked from commit ba38688)
Signed-off-by: Honza Pokorny <honza@redhat.com>