glogging._get_user() fails to catch malformed authorization header

[jeiting]

I have a client that abuses the HTTP authorization header, passing "Basic (token)" where token is a random alphanumeric string. This causes the gunicorn logger to choke because, although it is a valid base64 string of data, it does not represent base64 encoded utf-8 string.

[2018-01-15 18:23:06 +0000] [15] [ERROR] Error handling request
Traceback (most recent call last):
  File "/app/.heroku/python/lib/python3.6/site-packages/gunicorn/glogging.py", line 269, in 
    'u': self._get_user(environ) or '-',
  File "/app/.heroku/python/lib/python3.6/site-packages/gunicorn/glogging.py", line 446, in 
    auth = auth.decode('utf-8')
UnicodeDecodeError: 'utf-8' codec can't decode byte 0x8b in position 0: invalid start byte

You can reproduce it with the following line:
base64.b64decode("ixsTtkKzIpVTncfQjbBcnoRNoDfbnaXG".encode('utf-8')).decode('utf-8')

I don't expect _get_user() to work properly in this case but it would be good for it to catch the error. Currently the code is in a try/catch but it is only handling TypeError, and binascii.Error, it should also catch and handle UnicodeDecodeError. I'm working on a PR now, should be ready shortly.