Revisit API status

[vishalbollu]

Improve the observability of the cortex statuses to help users clearly understand the statuses of their API pods without having to drop down to kubectl.

The API status has at least two confusing statuses:

• updating when the API is initializing and updating when there is a new version has the same status except that you will get 500s in the former and you will get actual responses in the latter
• compute unavailable is a pod that has been stuck in the pending status for 15 minutes. This can happen for other reasons such as API server is not responsive.

Also consider distinguishing between pending / pulling image / initializing (readiness check failing)

Also consider identifying if a new node has been requested (or was not able to be), if the new node is spinning up (not joined the cluster), and if the new node has joined and is initializing (the pod is not yet running on it)