Deploy HF docker image before HF maven artifact in the nuxeo-hf job

Sometimes copying the HF docker image from docker.platform.dev.nuxeo.com to docker-internal.packages.nuxeo.com fails.
This happens right after deploying the corresponding maven artifact to packages on the maven-team-platform-private repository.

When this happens, we end up in a broken state: we need to delete the maven artifact manually before relaunching the job. (Or we will get an error 400 when trying to redeploy the already existing artifact)

Inverting the deployment order would avoid to end in that broken state.
Allowing a retry while waiting for eventual further investigation. (dropped, see comment below)

As this only happens on 2021, the retry fallback will not be ported on 2023. This will allow the following:

• if there is a HF size threshold, it will show when 2023 starts to fail
• if this is a skopeo version issue, it will remain on 2021 and disappear with it

The stack trace when the problem happens:

[2024-02-14T02:04:35.469Z] Copying blob sha256:db66307d8564d71a7d6c5de69924172db191c90985365019e66fec24247d713e
[2024-02-14T02:07:42.017Z] time="2024-02-14T02:07:29Z" level=fatal msg="writing blob: Patch \"https://docker-internal.packages.nuxeo.com/v2/nuxeo/nuxeo-hf/blobs/uploads/d46570e0-e558-4d34-bec9-6bb990afc763\": happened during read: Digest did not match, expected sha256:469b27d5a67d284b9bf8f9e771e8647e51b341cd5664771864a7f04edf1e7f81, got sha256:2b9ef0eb3572c4f8f11505c83c3ee39684fdd4cfa031e5c10168a7dbaaa088b8"
script returned exit code 1