Add Matcha-TTS

[alexhartl]

I've recently implemented the Matcha-TTS model for Text-to-Speech in Keras, together with a MelGAN vocoder. Matcha-TTS is a flow-matching model presented at ICASSP 2024, likely one of the best open TTS models out there. If you are interested, I would adapt it to KerasHub style and prepare a PR.

Paper: https://arxiv.org/abs/2309.03199
Original PyTorch implementation: https://github.com/shivammehta25/Matcha-TTS
Demos for the original implementation: https://shivammehta25.github.io/Matcha-TTS/