Use bytes or %r to handle paths and process output

[theotherjimmy]

Paths are not ASCII, paths are not Unicode. Paths are bytes.
This PR handles paths like they are bytes, well at least the
current working directory. Handling paths and process output as
if they are unicode will introduce errors when paths are not
valid unicode. I tested this by creating a directory named

�abc\uXXXX

Which is not valid unicode, and tricks most python string
decoders into throwing an exception.

Fixes #710, maybe

[screamerbg]

Looks great! Thanks @theotherjimmy. You might want @toyowata to check this with his Japanese dev environment before merging.

[theotherjimmy]

I should get it working with python3 as well.

[theotherjimmy]

This is very involved. Paths and other bytes being treated interchangeably with strings is everywhere in this code base.

[theotherjimmy]

Closing in favor of #767