fix: fix for comparing semver for enterprise versions

[pmalek]

What this PR does / why we need it:

This fixes an issue where an enterprise version (like e.g. 3.4.0.0) wouldn't be allowed to be provided.

Which issue this PR fixes

Fixes #856

Special notes for your reviewer:

Checklist

[Place an '[x]' (no spaces) in all applicable fields. Please remove unrelated fields.]

• PR is based off the current tip of the main branch.
• Changes are documented under the "Unreleased" header in CHANGELOG.md
• New or modified sections of values.yaml are documented in the README.md
• Commits follow the Kong commit message guidelines

[rainest]

My kingdom for Enterprise using standard semver build tags exclusively for its extra stuff so it at least parses and sticks its mystery values in the designated mystery value box.

What's the intended effect of the change? Regex inscrutability on top of the normal template mess inscrutability will be hard to read in the future, so we should have an explanatory comment inline explaining how this intends to parse input.

We presumably need ^ anchors, correct? What happens if we get a 4.3.4.1 where the tail is a valid semver version we'd compare against?

[randmonkey]

What if the latest tag is used (as used in KIC nightly integration tests)?

[pmalek]

We presumably need ^ anchors, correct? What happens if we get a 4.3.4.1 where the tail is a valid semver version we'd compare against?

Yes, added that.

I've simplified the change to just extract the first 3 segments if those are detected, otherwise fallback to old behaviour. This will fix the 4 segment versions and leave the old behaviour otherwise.

[pmalek]

Not sure why the tests are failing 😠 https://github.com/Kong/charts/actions/runs/5831706278/job/15815656070?pr=857#step:10:2018 the pods deploy successfully 🤔

[rainest]

Problem diagnosis: chart-testing is terrible at reporting issues. --helm-extra-args "--debug" will make Helm explain its reasoning, because for some reason it doesn't actually explain them on failure.

Do not use chart-testing's --debug in most cases, since it suppresses Helm debug output for some reason.

Deleting the other CI files temporarily makes the output less of a mess.

Problem cause: chart-testing's KIND does not use any LB provider, and Helm will block installs if a spawned LB never gets addresses. All the CI entries need proxy.type: NodePort.