Remove usage of gofuzz

[richardcase]

CAPI (and providers via importing the CAPI package) uses gofuzz for fuzzy testing api conversions via the conversion package.

The gofuzz project has been archived since Jan 2024 so it's not receiving any updates. It's probably worth us considering moving away from using this package to something that is maintained or in-tree.

There is a lot of benefit from fuzzy testing the api conversions so we shouldn't remove these types of tests.

Upstream Kubernetes has marked this as unwanted and suggests randfill be used.

Similar to #9548.

/area testing
/kind cleanup

[chrischdi]

From the randfill docs

This is a fork of github.com/google/gofuzz, which was archived.

NOTE: This repo is supported only for use within Kubernetes. It is not our intention to support general use. That said, if it works for you, that's great! If you have a problem, please feel free to file an issue, but be aware that it may not be a priority for us to fix it unless it is affecting Kubernetes. PRs are welcome, within reason.

So looks like that's the way to go if the library works for us as a replacement.

[sivchari]

At first glance, we're depending on fuzzing features provided by apimachinery. So if we migrate from gofuzz to randfill, we also remove the apis/meta/fuzzer and apitesting/fuzzer dependecy.

[sbueringer]

Didn't have time to look into the details, but +1 in general

[sivchari]

/assign
It's possible to switch to randfill, but the switching requires Go1.24.0 >=, so we can complete it after #12128 is merged

[chrischdi]

/priority important-soon
/triage accepted

Will be part of controller-runtime bump:

• ⚠️ Bump to controller-runtime v0.21 / controller-tools v0.18 / k8s.io/* v0.33 / move to randfill #12191