This was only half-applied from #213 (closed), and there still a few things missing. This continues on from that (though in a slightly different way) and also fixes the code in the regtest directory. Note though that there's not really any documentation for it, so I only tested that it worked superficially (but most of the conversion was automated anyway.)
As for the difference between 2 and 3, I'm pretty sure this is because newer versions use newer Unicode tables. For example, running just Python 2.7 updates the table length from 5143 to 5516 (UCD 5.1.0). With Python 3.6, the table is 5722 entries (UCD 9). Python 3.7 has UCD 11, so probably there would be even more entries, but I did not try it out.
I guess that poppler/UnicodeDecompTables.h should probably be updated to take advantage of the newer tables, but I did not do that here.
Closes #213 (closed).