Create a CI harness to auto-detect and mark archetype drift when changes break it #4100

We have a wonderful repo, file and function classification system based on k-means clustering on file metrics sans extension, so we can cluster these extractions by structure and not language. But we've also been rapidly improving the engine, quality of detection, etc. I'm worried that we aren't correctly monitoring drift in that system when we grow and potentially are reporting incorrect archetype classifications.

This is an important gap we should explicitly monitor and put into the ML archetype meta data if it was validated or not. So the system know's it's trustworthy for that version.