Live text-to-speech streaming

Test environment for streaming TTS: text is split sentence by sentence, each sentence is synthesized and streamed back the moment it's ready, and played gaplessly as it arrives — you don't wait for the whole thing to finish generating.