Failed Compute Nodes Show Instances as Running instead of Unkown.

[Smithx10]

We cannot reschedule an instance based off the state of the Compute Node being in an unknown state. I don't know how much of an impact this is in reality, but in the event a CN goes down... It's instances should reflect that state.

Currently we will just respect the normal backoff window and attempt to restart the instance 3 times. If those restarts do not occur, we will reschedule.

There is a issue filed on sdc-vmapi.
TritonDataCenter/sdc-vmapi#35