Add support for German TTS with Thorsten dataset

[monatis]

Closes #209.
It supports both training and inference with Thorsten dataset, adding link to inference notebook in readme.md with pretrained tacotron2 and multiband_melgan models in German.

[dathudeptrai]

@monatis awesome work, I really appreciate your work. BTW, can we redistribute your pretrained weights in our drive ?

[monatis]

@dathudeptrai thanks! 🙏
Sure, you can redistribute them. I will test it out with some more training steps and I'll release it if I make improvement in the quality of synthesis. I'll also add TF Lite inference example to the Colab notebook 🚀

[dathudeptrai]

@monatis thank you very much :D

[thorstenMueller]

Thanks @monatis and @dathudeptrai for your great effort on training on my dataset 🥇 . I appreciate it a lot.
@monatis, when you've new samples with more training steps please let me know and i'd be happy to update them on my comparison page.