Use JS glob + ignore for the cache glob step

[nekowinston]

The current find ${{ inputs.directory }} -regex '.*\.[jt]sx*' -exec deno cache {} \; stuff is pretty... gross.
I'd like to make picking the files which deno caches more customizable and intuitive.

Ideally, this would be handled by the deno binary itself; e.g.:

Repo layout:

/
├── test/
│   ├── basic/
│   │   └── main.ts*
│   ├── customized/
│   │   ├── deno.json
│   │   ├── deno.lock
│   │   └── main.ts
│   └── problem-matcher/
│       ├── deno.json
│       ├── deno.lock
│       ├── foo.ts
│       ├── foo_test.ts
│       └── main.ts
├── .denoignore
└── deno.json

.denoignore:

test/problem-matcher
!test/problem-matcher/foo.ts

$ deno cache
-> caching test/basic/main.ts
-> caching test/customized/main.ts
-> caching test/problem-matcher/foo.ts

---

Working around the fact that this isn't supported by the binary, we could create a script that works in both Node (for actions) and Deno (for the dev), which effectively does this. The dev could add this script to the deno.json like

{
  "tasks": {
    "cache": "deno run -A https://esm.sh/gh/nekowinston/setup-deno/cache.ts"
  }
}

Which then calls deno cache on each file in the glob pattern, storing it in deno.lock.
The Node script would have the same files in its match, with improved customizability compared to just a single action input for a directory.