perf(ci): speed up Danger by switching from Dockerfile builds to GHCR…

[yoshi-taka]

CI speed improvement

https://github.com/jest-community/eslint-plugin-jest/actions/workflows/lint.yml

• GHCR images (ghcr.io/danger/danger-js) are prebuilt and cached on GitHub-hosted runners
• Completely removed Dockerfile-based builds

[G-Rath]

Could you explain this change a bit more, ideally with some linkage to official documentation? My expectation is that GH already uses caching when appropriate for docker images, and doesn't build them every time.

I'm also not sure if this is supported by Renovate

[yoshi-taka]

As you can see below, this PR runs noticeably faster, dropping from about one minute to twenty seconds compared with many other PRs.

If we look at what the previous action was doing, the below screenshot shows that it was performing a full Docker build.

e.g. https://github.com/jest-community/eslint-plugin-jest/actions/runs/19342384594/job/55333811424#step:2:1

GitHub Actions has a behavior that many people are not aware of:
when you reference a repository with uses:, Actions builds a Docker image from that repository’s Dockerfile.

[yoshi-taka]

Renovate is able to update pinned Docker digests, as documented here: https://docs.renovatebot.com/docker/#digest-pinning

[G-Rath]

I understand it aims to speed up CI, but it's not clear to me if this is official, stable, trust worthy, etc which is why ideally I'd like some docs or a blog post

[yoshi-taka]

danger-js has been publishing images to GHCR for quite some time.
https://github.com/danger/danger-js/pkgs/container/danger-js

The code that handles the publishing is here:
https://github.com/danger/danger-js/blob/main/.github/workflows/publish_package.yml

[yoshi-taka]

This is the closest official reference I could find: https://github.com/danger/danger/blob/master/CHANGELOG.md#932