Currently we use Unified CI for tagging releases based on version.json, but we'd like to minimize the effort for cutting releases so that we can cut them more frequently.

So I propose we instead cut releases using https://github.com/pl-strflt/changelog-driven-release. Basically the workflow would look like:

• Accumulate changes in the [Unreleased] section of the CHANGELOG
  • These should be added as commits are added, and the CHANGELOG update should happen as part of the PR for a feature
    • CI enforcement of this is tracked here: Enforce changelog updates in CI #267
• When ready to release, change [Unreleased] to [vX.Y.Z] YYYY-mm-dd, which is then detected by CI as a release and results in a new tag and a GitHub release with the changelog contents in the release notes.

Other requirements:

• We currently run Kubo tests for each release, until enough are moved into Boxo, so we need to run releases on a release-vX.Y.Z branch and merge into release which should trigger the release. This way we don't have to "freeze" main while we do a release. Once the release is complete, merge release into main (which should NOT trigger a release). This is different than most folks, who just merge to main which triggers a release based on the tip of main...we do NOT want this behavior.
• We love all the checks that Unified CI does for releases and would like to keep them (release checker workflow, which runs gorelease et al to verify the release version, call out breaking changes, etc.)