Always decode proc_output from UTF-8 on Python 2

[petedmarsh]

proc_output only decoded from UTF-8 if the input string was not an instance of str. This seems to be for Python >=3.5 as the the type of stdout returned by subprocess.run(...) is bytes not str. In Python 2 a series of bytes is represented by the str type and not a separate bytes type, so the output of subprocess.Popen will be of type str but this is really a series of bytes representing a string in some encoding. Fortunately in Python 2 bytes is an alias for str so proc_output can change to decode when it receives bytes.

For complete correctness proc_output should decode with the encoding of the bytes in the output these suggest that the decoding to use is the result of os.getfilesystemencoding():

https://github.com/pyinstaller/pyinstaller/issues/1325 https://bugs.python.org/issue19846

However, other parts of pipsi assume UTF-8 encoding (it is a safe default) so proc_output has been left to assume UTF-8.

Fixes #133

[petedmarsh]

The build fails for Python 3.3 which is an issue other PRs have had too, I don't think it;'s related to this change.