GustavMauler

Tested quickly with the mic example. The Stream object created by Models manage themselves in deepspeech now, rather than being managed by the Model (for stopping etc.) Additionally deepspeech now...