Source PR: https://github.com/containers/automation_images/pull/388

The intention there was to force zstd:chunked compression on all images in the CI image cache. (Describing said cache is beyond the scope of this description).

Once podman and skopeo fully support zstd:chunked, this PR should be revisited. Questions to ponder: should all images be zstd:chunked, or only a subset? If subset: random, or hand-picked? And, do you need to take v2s1-v2s2-etc into consideration?