If we find stuff that's "identical" then we should really compare to see if it's ACTUALLY identical or not, or else this can make some things REALLY weird
This is the same diffing that is used by Romba to output its diffs. It is currently untested. Basically, this will take a set of base DATs as an input, and then diffs all of the individual DATs against that combined base and then outputs accordingly.
Previously, sets that had multiple items would trigger issues because of shared information in the items, mostly the Machine parts. By making all of it a clone, it takes a little more memory but ends up resolving all issues in desc-to-name mapping and replacement.
