py2/py3 - always use string object in re.sub

[pztrick]

Needs to be tested on actual python2 project/environment for compatibility.

Python3 returns a bytes object which re.sub cries about. I also was not sure if there was reason for usage of ascii encoding so I switched to utf-8.

[vilcans]

I think the ascii encoding is needed to make sure special characters are encoded as XML entities instead of being kept, but I can't check it right now.