Automatic vale improvements

[gingerwizard]

We need a way to be able to retrospectively apply vale checks across the entire corpus. To make this more efficient and rather than check these manually, proposing the following process:

1. We add a .check file to a folder (empty means all md files i.e. **/*.md[x], but it can contain a glob to limit files)
2. Raise a PR
3. A CI/CD workflow finds all files beneath this folder recursively.
4. It invokes vale rules for these files
5. The ouput is parsed and the lines read and passed to an LLM with the value rule - asking for correction
6. The correction is output into the Github worker stream in a format which causes a suggestion to be proposed.
7. The user accepts or declines the suggestions
8. Commits the PR

The benefit here is we can do folders incrementally, AI does all the work but we review changes.

We need to establish if (6) is possible.