Fix bytes/string issue when using chunked transfer encoding

[mstojcevich]

When chunked transfer encoding is used (the default behavior for responses to HTTP 1.1 requests) an exception gets raised because transform_chunk returns an str, which then gets appended to headers in flush, which is a bytestring, so you get a traceback like this:

  File "/home/marcus/......../twisted/internet/defer.py", line 654, in _runCallbacks
    current.result = callback(current.result, *args, **kw)
  File "/home/marcus/src/cyclone/cyclone/web.py", line 1159, in _execute_failure
    return self._handle_request_exception(err)
  File "/home/marcus/src/cyclone/cyclone/web.py", line 1211, in _handle_request_exception
    self.send_error(500, exception=e)
  File "/home/marcus/src/cyclone/cyclone/web.py", line 766, in send_error
    self.finish()
  File "/home/marcus/src/cyclone/cyclone/web.py", line 746, in finish
    self.flush(include_footers=True)
  File "/home/marcus/src/cyclone/cyclone/web.py", line 708, in flush
    self.request.write(headers + chunk)
builtins.TypeError: can't concat str to bytes

There was also another bug where ChunkedTransferEncoding.transform_chunk would include b' prefixes on the chunk size and chunk contents.

[coveralls]

Coverage remained the same at 0.0% when pulling 3132b59 on mstojcevich:cte-fix into fefdc51 on fiorix:master.