caching per-revision exp show output

[skshetry]

This equally applies to metrics/params and plots. We can cache the computed results from every revisions which might be faster than trying to cache the whole index.

At first, we could try caching results without any error but in the long run, we should be able to distinguish between retryable errors (eg: FileNotFoundError in cache and remotes) vs unretryable errors (eg: dvc.yaml being broken and since it's tied to an immutable commit, it's never going to be fixed).

[dberenbaum]

Should we break this issue into subtasks? I think the initial request of caching exp show output per revision is done, right?

[skshetry]

I think we can close this as this was implemented in #9069.

[dberenbaum]

Should we open other issues for these items?

This equally applies to metrics/params and plots.

At first, we could try caching results without any error but in the long run, we should be able to distinguish between retryable errors (eg: FileNotFoundError in cache and remotes) vs unretryable errors (eg: dvc.yaml being broken and since it's tied to an immutable commit, it's never going to be fixed).