Update OperatorSDK (1.27.0 -> 1.31.0) and Kubebuilder (go/v3 -> go/v4)

[phlg]

Work very much in progress.

This merge requests aims at updating Operator SDK, following the version-by-version guidelines documented here : https://sdk.operatorframework.io/docs/upgrading-sdk-version/v1.31.0/ ; this probably encompasses several PRs made by renovate, but they should (... I think ?) come as a coherent bundle "approved" by Operator SDK, instead of isolated PRs.

A first step was done with 5661ada ; this implemented the changes documented in aforementioned website starting from 1.28.0 all the way to 1.31.0 included, safe for two things :

• The 1.31.0 upgrade notes mention adding secret-watching capabilities to the role running the operator, which we did NOT implement. After digging (mostly through the release notes, see also the comment below) this only appears to be needed for Helm-based operators, so we should be safe. Hence this won't be implemented at all.
• The 1.28.0 upgrade notes mention go/v3 being deprecated by kubebuilder. As such, we should aim to migrate to go/v4 (or maybe go/v4alpha, unsure about this), probably following this documentation : https://book.kubebuilder.io/migration/manually_migration_guide_gov3_to_gov4.html
  • This PR being a draft is mostly due to this remaining TODO.

For future reference, the actual release notes on the Operator SDK Github repo (https://github.com/operator-framework/operator-sdk/releases) appear to be of much better quality than the upgrade documentation on their website. For instance, the upgrade documentation for 1.30.0 appears to be a copy-pasted placeholder, while there are in fact a few actual changes visible in the 1.30.0 release note.

[phlg]

The image built with the first commit was deployed in-house, and didn't generate any error. This is a very rough testing, but still better than nothing I guess.

[phlg]

The changes in 5ac28ef through 1e96f19 are all related to upgrading kubebuilder from go/v3 to go/v4.

The image obtained after theses commits was again locally tested and did not seem to generate any error, so hopefully the update should be bug-free. And if not, it will be an argument in favor of spending some time to implement actual tests in this operator.