excluding checks in admin `/ready` endpoint

[steve-chavez]

We could take the Kubernetes Individual Health Checks as reference, something like /ready/schema-cache-reloaded. Or if we want to check if we're serving requests regardless of schema reload success, we could verify /ready?exclude=schema-cache-reloaded.

Here's another similar implementation example I stumbled upon for reference: tikv/pd#9846

Originally posted by @laurenceisla in #4869 (comment)

[steve-chavez]

Motivation on:

IMO it should be an operator's decision which one to check in readiness probe. Hence I would suggest two different URIs: one would return 5xx on "no schema cache", the second one - on "no schema cache or known stale schema cache". It does not matter if these URIs are different paths or some different query parameters - don't want to bikeshed about this.
#4869 (comment)

I think we should use the /ready?exclude=... endpoint for this. Edit: on second thought /ready/schema-cache-reloaded is more direct, so I wouldn't be opposed either. (both could also be done)

[wolfgangwalther]

/ready/schema-cache-reloaded

If we're going this way, then let's please skip the -reloaded suffix, it doesn't make any sense, imho. /ready/schema-cache will tell you whether the schema cache is ready.