Version LiNT formula

LiNT uses the difficulty score and level that are calculated in t-scan, but also stores the level boundaries for some of its own visuals and explanatory texts. This results in some wonkiness when the formula is updated in t-scan but not (yet) in LiNT.

A possible solution is that t-scan could
- Associate version numbers with difficulty formulas.
- Let the user pick the version to use when starting the scan. Should default to the latest version, of course.

Then LiNT would continue to request the "old" formula until we update it.

This would also be a neat feature for researchers - while updates to the formula are always minor, it is nice for comparison if you can make sure that everything is scanned with the same formula.

No source file, test, or entry point is named. Start by locating the difficulty-formula implementation and the scan-start configuration, then trace how LiNT consumes formula results and level boundaries; done means a versioned formula can be selected with the latest as the default and older consumers can request a fixed version.

Written by the indexing model from the issue text.