Recursively read files in directory/path into JSON output

Nobody has claimed this yet.

I do have a GitHub repository for a Mozilla Firefox add-on (WebExtension). It does have changelog data listed in the repository in the following structure (see live in https://github.com/rugk/mastodon-simplified-federation/tree/main/assets/texts):

```shell
$ ls -a assets/texts
. .. de en fr pt_BR
$ cat assets/texts/en/Changelog/2.2.html

• New: The translation for Dutch has been added, see #118.


• Fixed:: Fixed problem sometimes duplicating tabs when interacting with Mastodon instances, see #117.

More information on GitHub.
```

So for the given example, if I have a fixed version number (which is possible to find out) I want an output like this:
```yaml
{
"en": "",
"de": ""
# etc.
}
```

So, the structure is always the same and I want to read all the content in a GitHub action and put it into a JSON format.

It should then be passed as a changelog/release notes into the following GitHub Action:
```yaml
steps:
- uses: wdzeng/firefox-addon@v1.1.0-alpha.0
with:
addon-guid: your-addon's-guid
xpi-path: your-addon.zip
self-hosted: false
release-notes: {"en-US":"foo"}
jwt-issuer: ${{ secrets.FIREFOX_JWT_ISSUER }}
jwt-secret: ${{ secrets.FIREFOX_JWT_SECRET }}
```

https://github.com/wdzeng/firefox-addon/issues/2#issuecomment-2040386656

## Idea

A real action to read multiple files would be great and simplify stuff here.

In my case [I asked ChatGPT and got a quite complicated script](https://chat.openai.com/share/26f3c54e-3d38-476e-bb87-9cea84e0ce97).

No implementation files, tests, or entry points are named. Start by inspecting the repository's GitHub Action inputs and output handling, then trace how file contents are currently read. Done means recursively reading the requested directory structure and producing JSON keyed by directory or locale with each file's content.

Written by the indexing model from the issue text.