Major Description
During text extraction, font size / descriptor / encoding attributes are accessed repeatedly in order to do positional calculations and byte-character conversions.
The current code has several accessors for these things that redo rather slow calculations each time - even thought the font object state is not changed.
The results of these calculations should be persisted in instance fields once calculated. This greatly improves performance.
I'll attach new versions of PDFont, PDFontDescriptorDictionary and PDSimpleFont that have these tweaks.