clipper-registry/cache

This is a GitHub Action that speeds up continuous integration builds by caching directories using a service called Clipper. It mounts a folder from your build using a special filesystem, so large files only get downloaded when something actually opens them, instead of pulling the whole cache down at the start of every job. To use it, you first need a free Clipper account and a token with permission to push, pull, and create data, which you save as a secret in your GitHub repository. Then you add the action to your workflow file, telling it which local folder to cache, which remote location to store it under, and a key name. The action automatically appends your current git branch name to that key, so it tries to find a matching cache for your exact branch first, then falls back to the cache from your base branch, and finally just starts fresh if nothing matches. When your job finishes successfully, any changed files get pushed back up for next time, and you can optionally have it push even when the job fails. A handful of settings let you tune how it works: which file patterns to split into smaller packs, whether to use content aware chunking when pushing changes, how many files to transfer at once, and which version of the underlying Clipper command line tool to use. Right now it only supports Linux based CI runners, with macOS and Windows support planned for later. There is also a Rust specific incremental cache setup included, with support for the ccache build tool listed as coming soon. The core idea behind Clipper itself is that traditional caching tools bundle everything into a single tarball, which is slow and wasteful when your build only actually needs a fraction of what got cached. Clipper instead indexes the individual files in a directory and only uploads the ones that have actually changed, while also letting different caches share identical files between them to save even more space and time.