GHA jobs fail because of too many requests

[jakirkham]

Have started seeing GHA jobs fail because of too many request (for example:

#16 ERROR: toomanyrequests: 
------
 > pushing rapidsai/staging:miniforge-cuda-71-cuda12.0.1-base-rockylinux8-py3.9-arm64 with docker:

Should we consider some changes here to mange this?

1. Reduce number of images built
2. Add a shared docker pull step to minimize pulls and reuse across builds
3. Combine all pushes into a single command at the end
4. Minimize how many jobs are run at once
5. Other ideas?

Note: This happens with ci-imgs too ( rapidsai/ci-imgs#162 )

[ajschmidt8]

The quickest solution is probably just to adjust the number of concurrent jobs that are allowed in the matrix workflow:

https://docs.github.com/en/actions/using-jobs/using-a-matrix-for-your-jobs#defining-the-maximum-number-of-concurrent-jobs

[bdice]

Closing in favor of rapidsai/ci-imgs#162, now that miniforge-cuda is being merged into ci-imgs (rapidsai/ci-imgs#167).