Caching MISE_CACHE_DIR or not in CI

[scop]

I wonder if it's recommended to cache the mise cache dir in CI or not? There are some mixed signals about it

• the recipe in https://mise.jdx.dev/continuous-integration.html#gitlab-ci does cache it
• it appears that jdx/mise-action does not, https://github.com/jdx/mise-action/blob/main/src/cache-save.ts

[jdx]

I think there probably isn't much value in doing so which is why I didn't do it in the action—it would just use up space unnecessarily. That may not be true in use-cases I'm forgetting about or may not be true in the future. I can't think of a reason other than space for not adding it though.

[hverlin]

We can remove it from the docs if it's not useful.

I think for gitlab-ci (and likely other CI), it might be better to use the ./bin/mise from https://mise.jdx.dev/cli/generate/bootstrap.html (using mise generate bootstrap -l -w).

This way the cache is conveniently always located at the root of the repo, and this works better with GitLab at least. This is what I do now:

.mise-cache: &mise-cache
  key:
    prefix: mise-
    files: ["mise.toml", "./bin/mise"]
  paths:
    - .mise/cache
    - .mise/installs
    - .mise/mise-2025.1.0

my-job:
  stage: build
  image:
      name: '<some-debian-based-image>'
  script:
    - ./bin/mise run build
  cache:
    - <<: *mise-cache
      policy: pull-push

However, probably we need to remove the experimental status of the bootstrap command before we document it this way?
I think it might also be nice if we can the version from ./mise/mise-<version> so one does not need to update it everytime they update ./bin/mise

[jdx]

the other thing about the cache is most of it expires in 1 hour or a day, so saving it in a CI context is not likely to be used again