pdf.py: minimal but hard conversion from Python2 to Python3

[thielema]

abuse latin1 encoding for working with Unicode strings as representation for a bytestream

Maybe better use https://github.com/agl/jbig2enc/pull/66/