feat(k8s): add Helm Chart

[apricote]

What this PR does

• Adds a helm chart for the hcloud-csi driver (Closes Helm chart #108 feat: Publish Helm Chart #369)
• Remove the old kustomization based manifests in favor of the Helm Chart
• Use the Helm Chart to render the static manifests that are currently used for install
• Publish the Helm Chart to https://github.com/hetznercloud/helm-charts

Reviewer Notes

This replaces #398 as I did not have push permissions for the branch.

I recommend reviewing by commit, to reduce the noise. This is still a huge PR and if requested I can split it up into multiple.

A lot of the scripts and goreleaser config was shamelessly stolen from hcloud-cloud-controller-manager.

[jooola]

I did not finish my review, but here is already a few points:

• Could we add tests for this ? The templating mixed with yaml feels a bit messy and I fear that it might hide a bunch of bugs. What about generating a set of charts with different .Values and compare it to some snapshots we keep in the repo?
• Could we reduce the complexity by writing good documentation instead of making the chart overly complex and handling every last field (e.g. not build the image field) ?
• Some files have an inconsistent naming, for example pbd.yaml or extra-list.yaml. I would also be in favor of using kebab case filenames.
• Should we introduce a values.schema.json file to validate the values files ?

[jooola]

Why not release 2.5.0 right after this is merged ?

[apricote]

I did not want to force our hands one this. I was planning on deploying the Chart to my personal cluster before creating a release, but I can also do that from this PR.

[apricote]

Could we add tests for this ? The templating mixed with yaml feels a bit messy and I fear that it might hide a bunch of bugs. What about generating a set of charts with different .Values and compare it to some snapshots we keep in the repo?

For sure, we already do that with the existing deployment manifest. I added two additional snapshots: default values & production example.

Some files have an inconsistent naming, for example pbd.yaml or extra-list.yaml. I would also be in favor of using kebab case filenames.

Removed extra-list.yaml and the associated functionality (deploying random resources?). Renamed pdb.yaml. Not so sure about the kebap-case, never seen that in a helm chart :D

Should we introduce a values.schema.json file to validate the values files ?

Sounds good :) In the latest push.

[jooola]

Nice!

[apricote]

I added an additional snapshot test that defines all optional values to make sure everything works. While doing that I found a bunch of features that did not work properly and fixed those.

I still want to do some manual testing before merging.

[apricote]

Manual testing went well. I also updated my personal cluster to use the chart from this branch and everything worked as expected :)

[jooola]

Great effort to build the snapshots tests! 💯