Merge change 6789 into donut

* changes:
  Add support in TextToSpeech for an audio stream type param and utterance ID specified as a hashmap param in the synthesis calls. Fix a bug where the cached parameters were not passed to the service when synthesizing to a file.