nova-compute charm workload status should block on inclusion in hypervisor list

[Ryan Beisner]

When the nova-compute charm is deployed, it is possible for the charm to represent a READY/OK workload state when in fact the compute hypervisor is not registered with nova successfully.

We should change that. The charm application and unit workload states should block until we confirm that the compute node is successfully registered with nova as a hypervisor.

[Ryan Beisner (1chb1n)]

Reference, where a compute node fails to register as a hypervisor, yet the Juju status is green:
https://bugs.launchpad.net/charm-nova-compute/+bug/1771662

The rest of that bug is unrelated, but it is a good example for this bug.

[James Page (james-page)]

To be honest, if the compute service fails to register correctly, should it not shutdown? Then service state is an accurate representation of ready-ness.