This change will add support for Kokoro TTS with txtai's TextToSpeech pipeline. It will use the ONNX model to be consistent with other supported TTS models.

Models should also be added for the INT8 and FP16 quants provided here: https://github.com/taylorchu/kokoro-onnx/releases/tag/v0.1.0

This model will require an update to the ttstokenizer project to support IPA phoenemes. We don't want to depend on espeak-ng given it's GPL licensed.