OPRUN-3458,OPRUN-3459: add policy enforcement for bundle metadata format based on OCP versions

[grokspawn]

This update augments the fbc-validation task to enforce bundle metadata formats for incoming fragments based on the destination catalog's OCP release.

For OCP releases before 4.17, this will ensure that all bundles use only the olm.bundle.object object.
For OCP releases at/after 4.17, this will ensure that all bundles use only the olm.csv.metadata object.

[jsztuka]

lgtm, +1 for discussion on failure format.

[grokspawn]

I'm not clear on how to kick the CI.... I see that the CI is not running because

User grokspawn is not allowed to trigger CI via pull_request on this repo.

Help?

[arewm]

/ok-to-test

[grokspawn]

🎉 Now it looks like there is an additional required action. I see an approval. Is there something I must do?

[arewm]

The Checkton check is showing violations in your task changes.

The trusted artifact variant check is likely flagging issues due to your change to the README generator. I know that we have gone back and forth on it, but you can back out that change if you don't want to resolve the differences. That change can be proposed later.

[grokspawn]

The Checkton check is showing violations in your task changes.

The trusted artifact variant check is likely flagging issues due to your change to the README generator. I know that we have gone back and forth on it, but you can back out that change if you don't want to resolve the differences. That change can be proposed later.

I understand what the tasks are communicating. The tasks aren't identified by the CI flow as required (and the checkton check didn't exist until 8/30).

To help folks contribute to this repo, it would be really helpful if

1. the set of required checks would reflect the merge check minimums; and
2. PRs that introduce checks should also vet impacts to existing code (and IMO create a PR to address, so that other contributors just have to merge/rebase to update their efforts).

[hongweiliu17]

The issue from #konflux-users might be caused by this commit, see analysis
The BASE_IMAGE is like registry.redhat.io/openshift4/ose-operator-registry-rhel9@sha256:56e5c57b9d94bcadad6ff0cd47d157d35affc4bfd2c23a2ac948eee1cac3b8a3, not registry.redhat.io/openshift4/ose-operator-registry-rhel9:v4.16, so that the OCP major version and minor version can't be derived correctly.

[chmeliik]

1. the set of required checks would reflect the merge check minimums; and

I don't think we treat any of the existing checks as optional. PRs don't get merged if the checks are not green.

2. PRs that introduce checks should also vet impacts to existing code (and IMO create a PR to address, so that other contributors just have to merge/rebase to update their efforts).

I believe that's how we do it (checks have to be passing before PRs get merged)