Fixes#2730. Usernames containing non-ASCII characters (e.g. 'Émile')
can trigger a UnicodeDecodeError inside the requests library during
redirect handling. This exception is not a subclass of
requests.exceptions.RequestException, so it escaped all existing
except blocks in get_response() and crashed the program.
Added a catch for UnicodeError (parent of both UnicodeDecodeError and
UnicodeEncodeError) so these sites are gracefully skipped instead of
crashing the entire scan.
Added regression tests in tests/test_unicode.py.
salmanrajz