python version-specific codepaths in tests

[Gordon Tyler]

John Arbash Meinel wrote:
> We probably could turn this into a 'win32tools.py' helper, and use it as
> long as ctypes is present (the code I have is using pywin32, but it is
> usually pretty easy to convert into a ctypes helper.)

You know, I swear I've seen code like this elsewhere in bzr using 
ctypes. I'll see if I can dig it up.

> As for whether SIGINT is *needed*... It is nice to see that things clean
> up cleanly and quietly. It isn't a strict requirement.

So being able to test that "bzr: interrupted" is output to stderr on 
SIGINT is proof that things were cleaned up?

Ciao,
Gordon