Run all formatter tests periodically

[raxod502]

Sometimes e.g. #312 there will be upstream changes that break a formatter test, and we only find out the next time somebody touches it. A nice way to get ahead of these things without hurting development velocity too much would be to make a simple status dashboard which could be populated with the output of running the full test suite of all formatters, and seeing which ones worked. Such a test could be run monthly on Github Actions, or similar.

We'd need a bit of refactoring to make it possible to run the test suite and get structured results back even if some tests failed, and even if some formatter installations failed during the Docker build.