Parse http1.1 chunks to find json

Chunked data sent via HTTP1.1 by Twitter is not as straight forward to read,
chunks indicating size of tweets are present no matter what, so we find the
json start in the buffer and forget the rest for now.

(Second part of correction for userstream and soon all streams. Fix python-twitter-tools#190 python-twitter-tools#116)

twitter/stream.py

@@ -32,11 +32,16 @@ def __iter__(self):
         if not self.block or self.timeout:
             sock.setblocking(False)
         while True:
+            utf8_buf = self.buf.decode('utf8').lstrip()
+            pos = utf8_buf.find('{')
+            if pos != -1:
+                utf8_buf = utf8_buf[pos:]
+                self.buf = utf8_buf.encode('utf-8')
             try:
-                utf8_buf = self.buf.decode('utf8').lstrip()
                 res, ptr = self.decoder.raw_decode(utf8_buf)
                 self.buf = utf8_buf[ptr:].encode('utf8')
-                yield wrap_response(res, self.handle.headers)
+                if isinstance(res, dict):
+                    yield wrap_response(res, self.handle.headers)
                 self.timer = time.time()
                 continue
             except ValueError as e: