[kube-state-metrics] Fix autosharding configuration service

[Alja9]

What this PR does / why we need it

Enforce ClusterIP for autosharding, avoid headless service error.

Which issue this PR fixes

Autosharding get a Kubernetes error, example:

level=WARN msg="upgrade failed" name=my-kube-state-metrics error="Service \"my-kube-state-metrics\" is invalid: spec.clusterIPs[0]: Invalid value: \"None\": may not be set to 'None' for NodePort services"
Error: UPGRADE FAILED: Service "my-kube-state-metrics" is invalid: spec.clusterIPs[0]: Invalid value: "None": may not be set to 'None' for NodePort services

Instead of giving an error, how about to enforce it to the ClusterIP (auto-corrected, no Kubernetes error).

(optional, in fixes #<issue number>(, fixes #<issue_number>, ...) format, will close that issue when PR gets merged)

• fixes #

Special notes for your reviewer

Checklist

• DCO signed
• Chart Version bumped
• Title of the PR starts with chart name (e.g. [prometheus-couchdb-exporter])
• Add an helm unittest test case to cover this change.

[sebastiangaiser]

Nice fix, this resolves the invalid Service manifest you get with autosharding plus a NodePort/LoadBalancer service.type (headless + NodePort can't coexist), and it also corrects the metrics-port nodePort guard that was checking service.nodePort instead of selfMonitor.telemetryNodePort. A few small things:

• Two of the unit tests are named "fails when autosharding enabled with NodePort/LoadBalancer service type", but they don't fail, they assert the service is forced to headless ClusterIP. Could you please rename them to match (e.g. "forces headless ClusterIP even when service.type is NodePort")?
• The test file is missing a trailing newline.
• Needs a rebase on main (currently conflicting) and a re-bump to 8.3.1.