Aggregated Discovery should not cache on legacy fallback

[Jefftree]

What happened?

Aggregated discovery caches when fetching from legacy discovery (https://github.com/kubernetes/kubernetes/blob/1ba14700a5448f0cb54b9f5a79100eea71d05e65/staging/src/k8s.io/kube-aggregator/pkg/apiserver/handler_discovery.go#L295-L297).

This causes apiservices who share the same service key to fetch from the cache and returns an empty document when their group-version differs from the initial apiservice that loaded the cache.

What did you expect to happen?

See above

How can we reproduce it (as minimally and precisely as possible)?

See above

Anything else we need to know?

No response

Kubernetes version

1.26

Cloud provider

local

OS version

# On Linux:
$ cat /etc/os-release
# paste output here
$ uname -a
# paste output here

# On Windows:
C:\> wmic os get Caption, Version, BuildNumber, OSArchitecture
# paste output here

Install tools

Container runtime (CRI) and version (if applicable)

Related plugins (CNI, CSI, ...) and versions (if applicable)

[Jefftree]

/triage accepted

[Jefftree]

/sig api-machinery