The next thing I want to do is change the pattern-matching approach
entirely so as to not use RegExps, by first tokenizing the files and
then doing word-by-word pattern matching for the licenses. That should
improve performance by orders of magnitude.