Test adding an option to interactively train matches for specific projects.

[joxeankoret]

The idea is to train with good matches from a diffing session using a random forest or decision trees approach, to try to improve the quality of new matches based on previous matches.

Using ML techniques makes sense in some scenarios, like pattern matching. Hopefully, but this is something that I have to try, I can make a simplistic model (probably using SciKit Learn) to improve local matches based on previous matches.