test_stop_rescue_mode_logs_and_raises_errors breaks when previous_status is DEPLOYED

[Gavin Panella]

The test does:

  node = factory.make_Node(
      status=NODE_STATUS.RESCUE_MODE,
      previous_status=random.choice([
          NODE_STATUS.READY, NODE_STATUS.BROKEN, NODE_STATUS.DEPLOYED]))

then later on:

  self.patch(node, '_stop').side_effect = exception

before calling stop_rescue_mode.

However, stop_rescue_mode doesn't call _stop all the time:

  if self.previous_status in (NODE_STATUS.READY, NODE_STATUS.BROKEN):
      self._stop(user)
  elif self.previous_status == NODE_STATUS.DEPLOYED:
      self._power_cycle()

The test should patch _power_cycle when the previous_status is DEPLOYED.

This begs another question: what happens when previous_status is not one
of those three statuses? The answer may be "that can't happen", but it
probably will anyway. Perhaps MAAS should power cycle when
previous_status is DEPLOYED and stop in all other situations.