re does not handle re.ASCII properly

[slozier]

Currently, the re module just ignores the re.ASCII flag. This flag is used in quite a few places in the standard library.

This causes test_shlex to fail.

[slozier]

Causes a failure in test_argparse (CPython 3.5 version). The failure reduces to this:

import re
assert re.match(r'\s+', "\xa0", flags=re.ASCII) is None

[slozier]

While #814 addresses the specific issues noted above, I'm leaving the issue open since there's probably more work to be done.