fix(ci): setting all the github action jobs to run on the edera runners

[marly-edera]

The overarching changes here are to update all build systems to target the self hosted runners in the dogfood-cluster. Some notes:

• docker-setup.sh is left dangling, it does not work with the docker:dind sidecar container. The follow up plan for this is to configure our own copy of the image and inject it in the dogfood-cluster repo (well until that gets migrated into the infra repo)
• any build steps that were targeting ubuntu-latest have been moved to target the edera-24 runners in the Self Hosted group. The Self Hosted tag is load bearing, it is tied to the actions.github.com/runner-group-name annotation of the AutoscalingRunnerSet resource in the dogfood-cluster
• switched from pip to uv. uv is a bit less painful to work with imo, and I think it will make our lives easier going forward. I'm not married to this but I do recommend it. At some point I would recommend moving this from a GitHub Action step to just installing uv directly into the container image for speed, but it's not a huge problem at the moment. To that end I set up the configuration necessary to move from pip to uv and replaced all pip and naked python invocations with uv run.
• also switched to a crane setup step instead of an install script we manage

https://linear.app/edera/issue/OOC-79/adopt-dogfood-ci-runners-into-linux-kernel-oci-repo

[found-it]

The edera-24 runners are massive (24 CPUs, 96G), way overkill for things like digestabot. We should roll out a couple more runner scale sets with smaller runners (edera-2, edera-4, edera-8, edera-16) so we can use them here. Non-blocking for this PR though, just a note for future