returned tuple may be None.

    Examples of authority parsing:

    >>> _parse_proxy('proxy.example.com')
    (None, None, None, 'proxy.example.com')
    >>> _parse_proxy('proxy.example.com:3128')
    (None, None, None, 'proxy.example.com:3128')

    The authority component may optionally include userinfo (assumed to be
    username:password):

    >>> _parse_proxy('joe:password@proxy.example.com')
    (None, 'joe', 'password', 'proxy.example.com')
    >>> _parse_proxy('joe:password@proxy.example.com:3128')
    (None, 'joe', 'password', 'proxy.example.com:3128')

    Same examples, but with URLs instead:

    >>> _parse_proxy('http://proxy.example.com/')
    ('http', None, None, 'proxy.example.com')
    >>> _parse_proxy('http://proxy.example.com:3128/')
    ('http', None, None, 'proxy.example.com:3128')
    >>> _parse_proxy('http://joe:password@proxy.example.com/')
    ('http', 'joe', 'password', 'proxy.example.com')
    >>> _parse_proxy('http://joe:password@proxy.example.com:3128')
    ('http', 'joe', 'password', 'proxy.example.com:3128')

    Everything after the authority is ignored:

    >>> _parse_proxy('ftp://joe:password@proxy.example.com/rubbish:3128')
    ('ftp', 'joe', 'password', 'proxy.example.com')

    Test for no trailing '/' case:

    >>> _parse_proxy('http://joe:password@proxy.example.com')
    ('http', 'joe', 'password', 'proxy.example.com')

    r