Add functionality to restrict the disk space used by the Cache

[gbrls]

Hello @epwalsh, thank you for this project, it's very useful!
One issue that I have is that the cache sometimes takes up all of the disk space, for when I'm doing an operation that moved a lot of data around.
A solution for that would be to implement a maximum size which the cache can occupy on disk and a strategy to remove entries.

If you have some specific idea or direction on how this could be implemented I'd be happy to do it and write tests for it.
Cheers

[epwalsh]

Hey @gbrls I'd be happy to review a PR for this. I see a couple paths forward:

1. Add a function that needs to be called manually by the end users. It could take several options that determine which artifacts are removed, like a glob on resource names (URLs), a size cutoff, etc.
2. Add a setting that does removal automatically every time cached_path() is called.

These are not necessarily mutually exclusive, and maybe (1) is just a good first step to (2) anyway.

As far as cache removal options, I think it would make sense to have an LRU option, i.e. remove resources that were accessed less recently first. For that we could keep track of last_accessed time in the resource metadata here.

[gbrls]

Cool! I'll first send a PR to delete the compressed archive after it has been extracted, that'd already help with the storage issue.

[epwalsh]

Perfect