Fix bug in Docker image build process which caused tags to have improper manifests in DockerHub

[tdonohue]

References

• Fixes a bug in Setup the Docker GH Action for Matrix Building #2644 discovered after the merger

Description

Fix bug in Docker build process, where the manifests for the latest and latest-dist images on DockerHub were invalid.

The core issues was that in #2644, both the latest and latest-dist build processes were sharing the same artifact to store their digests. When the manifest was built for one of those tags, it was including every digest that had been built so far.

This behavior resulted in the latest tag in DockerHub having a manifest that included all the following:

• The latest image version for AMD64
• The latest-dist image version for AMD64
• The latest image version for ARM64

It also meant that the latest-dist tag ONLY included the longest running image:

• The latest-dist image for ARM64

The reason for this was a race condition where the ARM64 image took longer to build. At the point where the latest ARM64 image completed, both of the AMD64 images were already completed. Since they shared the same artifact, both AMD64 images WRONGLY ended up included in the same manifest.

This PR also does some minor cleanup in this docker.yml including:

• Updating all action step plugins to latest versions
• Reordering matrix parameters to list the arch first, so that step/jobs names in GitHub are clearer.
  • Previously names looked like dspace-angular (false, linux/amd64, ubuntu-latest)
  • Now, then names will look like dspace-angular (linux/amd64, ubuntu-latest, false)
  • This is just my preference, but it's also easier to tell steps apart as longer names get trimmed in the display. It moves the boolean for isPR to the end as it's not useful in the trimmed name.

Instructions for Reviewers

• This will be merged immediately if it passes PR CI validation. The full test requires merger.

[tdonohue]

This improves the Docker image builds, but still results in a failure for the latest-dist image. Finally figured out what I believe is causing the issue & it should be fixed in #2650

[tdonohue]

Ported to dspace-7_x in #2655