Pass through bytes in path and querystring

Per discussion on #303, bytes should pass through even if they aren't
UTF-8 encoded.

src/treq/client.py

@@ -157,9 +157,11 @@ def request(self, method, url, **kwargs):
         elif isinstance(url, EncodedURL):
             parsed_url = DecodedURL(url)
         elif isinstance(url, six.text_type):
-            parsed_url = DecodedURL.from_text(url)
+            # We use hyperlink in lazy mode so that users can pass arbitrary
+            # bytes in the path and querystring.
+            parsed_url = DecodedURL.from_text(url, lazy=True)
         else:
-            parsed_url = DecodedURL.from_text(url.decode('ascii'))
+            parsed_url = DecodedURL.from_text(url.decode('ascii'), lazy=True)
 
         # Join parameters provided in the URL
         # and the ones passed as argument.

src/treq/test/test_client.py

@@ -73,6 +73,23 @@ def test_request_uri_encodedurl(self):
             None,
         )
 
+    def test_request_uri_bytes_pass(self):
+        """
+        The URL parameter may contain path segments or querystring parameters
+        that are not valid UTF-8. These pass through.
+        """
+        # This URL is http://example.com/hello?who=you, but "hello", "who", and
+        # "you" are encoded as UTF-16. The particulars of the encoding aren't
+        # important; what matters is that those segments can't be decoded by
+        # Hyperlink's UTF-8 default.
+        self.client.request('GET', 'http://example.com/%FF%FEh%00e%00l%00l%00o%00?%FF%FEw%00h%00o%00=%FF%FEy%00o%00u%00')
+        self.agent.request.assert_called_once_with(
+            b'GET',
+            b'http://example.com/%FF%FEh%00e%00l%00l%00o%00?%FF%FEw%00h%00o%00=%FF%FEy%00o%00u%00',
+            Headers({b'accept-encoding': [b'gzip']}),
+            None,
+        )
+
     def test_request_uri_idn_params(self):
         """
         A URL that contains non-ASCII characters can be augmented with