Fix:[1852] Add TestK8APIEndpoints and run it across k8s v1.33-v1.37 in CI

[sameerforge]

What this PR does / why we need it:

Adds TestK8APIEndpoints, an e2e test verifying that after deploying kapp-controller:

• the aggregated APIService v1alpha1.data.packaging.carvel.dev becomes Available
• data.packaging.carvel.dev/v1alpha1 discovery exposes packages and packagemetadatas
• packaging.carvel.dev discovery exposes packageinstalls and packagerepositories
• kappctrl.k14s.io discovery exposes apps

This closes a gap where kapp-controller's CRD/aggregated-API registration had no dedicated test coverage, so a regression here (e.g. the aggregated APIService failing to become Available on a given apiserver version) could only be caught indirectly, if at all. This test runs on every PR via the existing "Kind Cluster E2E tests" run-tests job, alongside the rest of the e2e suite, against the default kind node image.

Adds a new k8s-api-endpoints.yml workflow, on a daily schedule plus workflow_dispatch, that runs TestK8APIEndpoints across the range of k8s minor versions kapp-controller supports. Compatibility drift from new k8s releases isn't a per-diff regression concern, so this runs separately from PRs rather than adding checks to every one, following the same schedule + workflow_dispatch idiom as trivy-scan.yml/dependency-updater.yml. The minors tested are derived from the k8s.io/client-go version pinned in go.mod (the last 4 released minors plus the next one), so the range shifts automatically as dependency-updater.yml bumps go.mod instead of being hand-maintained. The latest patch for each minor is resolved at run time from kubernetes/kubernetes's tags (via git ls-remote, avoiding GitHub REST API rate limits). The job then loops over the resolved versions sequentially in a single job, creating and tearing down one kind cluster per version. A failure on one version doesn't stop the others: the loop runs through every version regardless, reports each failing k8s version via ::error:: annotations, and only fails the job once all versions have been attempted.

Sample CI run for this workflow :
https://github.com/sameerforge/kapp-controller/actions/runs/34329006415/job/102392900710

Which issue(s) this PR fixes:

Fixes #1852

Does this PR introduce a user-facing change?

NONE

Additional Notes for your reviewer:

• The k8s-api-endpoints job installs kind/kubectl once via helm/kind-action@v1 (install_only: true, pinned to v0.33.0 because the action's current default, v0.31.0, can't parse the containerd config v4 format that newer kindest/node images ship — this was breaking kind load docker-image for k8s v1.34+), builds the kapp-controller image once, then loops: kind create cluster → kind load docker-image → kapp deploy → run TestK8APIEndpoints → kind delete cluster, per resolved version.
• run-tests (full e2e suite) is untouched and keeps testing against the default/latest kind node image.

Review Checklist:

• Follows the developer guidelines
• Relevant tests are added or updated
• Relevant docs in this repo added or updated
• Relevant carvel.dev docs added or updated in a separate PR and there's
  a link to that PR
• Code is at least as readable and maintainable as it was before this
  change

Additional documentation e.g., Proposal, usage docs, etc.:

[sameerforge]

Sample CI run for this workflow :
https://github.com/sameerforge/kapp-controller/actions/runs/34329006415/job/102392900710