Additional support needed about pdfrw HOT 2 CLOSED

Attached file with the changes supporting flate decompression with png 
predictor. Source from pypdf. By the way I already implemented support for xref 
streams and object streams. It works fine, but I need to cleanup a little bit. 
If you are interested, later I can send you my implementation.

P.S. Be aware with pypdf emulation. Library is full of bugs. I'm working with 
millions of pages in production printing company and once tried to generate 
content with pypdf. On the screen documents were ok, but after converting to 
postscript and printing about 40000 pages, I saw a lot of shit with fonts and 
graphics at random locations. With pdfrw output, at least for now, is quite 
good and performance is better than hummus pdfwriter (I used it with my own sip 
wrapper until I came to pdfrw).