Tidy up the downstream mirroring logic

Martin assured this isn't going to be going away anytime soon, so this is an attempt to make the logic a little less obfuscated to ease maintenance in the meantime. The triggering scripts have been deleted. If you squint hard enough, ensure_container.sh was really only setting environment variables on the container, adding artifact scripts, adding the Mesa tarball and doing modifications of entrypoints. This can all be done in the CI files now.

Env vars are setup using the ci-common scripts upstream. The artfacts are part of the debian-testing job, so they can be pulled in via dependencies. Along with artifacts, the Mesa tarball can be pulled from job dependencies, rather than via LAVA minio upload. We don't depend on the kernels from LAVA, so all that stuff was deleted.

All this is shipped to the DUT using job folder sync, as per the upstream MR mesa/mesa!14660 (merged)

Keeping the two approaches in sync is critical. Debugging this was a nightmare due to subtle differences in the two CI approaches. Due to how long the feedback loops are in Gitlab CI, you want to keep the differences between the mirroring and the upstream as small as possible. The other tidy up commits can be seen across the b2c-traces-runner branch, which keeps some downstream bits,

https://gitlab.freedesktop.org/tanty/mesa-valve-ci/-/commit/c637a57502e51cc19999135bc8dad6059e9501cc https://gitlab.freedesktop.org/tanty/mesa-valve-ci/-/commit/72255c3574a19f2154eb1d80647c1ec3c3f123de https://gitlab.freedesktop.org/tanty/mesa-valve-ci/-/commit/4fef0ef1eeb255b053f8849e1a9cb25403fe1ec8

And the mirroring helper, the new CI bits to make ensure_container redundant, and bits to make valvetraces run in the new env,

tanty/mesa-mirroring-helper@f248493b The following three should be squash really, but for reasons, they are not, tanty/mesa-mirroring-helper@c6962c85 tanty/mesa-mirroring-helper@20d69b94 tanty/mesa-mirroring-helper@5d1bd6a4