To help scaling nova-api and nova-metadata service in 18.0 we need to tune the liveness and readyness probes.

See the comments in the bug in the parent Epic for details

Acceptance criteria:

• k8s does not kill a busy nova-api or nova-metadata pod but instead it does not forward more traffic to them until the pod becomes less busy again
• k8s still kills the pod if it hangs for an extended period of time to give it a change to recover by recreating it
• the APITimeout parameter of these services are taken into consideration in the probe configuration. So if the customer configures a longer APITimeout for a service the k8s is also willing to wait longer before removes the pod from service or kills it.