utf-8 too and we don't end up mixing utf-8 characters (from %xx
translation) into a non utf-8 sequence.

same parameter. This is useful for APIs such as the LinkedIn
search API using multiple instances of the "facet" parameter.
Also updated get_normalized_parameters to handle the new
datatype returned by _split_url_string (list of tuples instead
of dict).
It doesn't break from_request but it doesn't make it possible
to use multiple instance of the sameparameters in from_request.
(as they are stored in a dict in that function)

Also added a new test_split_url_string test.