Towards v0.1.0: starting new dev updates

[r9y9]

I've just tagged a new release v0.0.3: release https://github.com/nnsvs/nnsvs/releases/tag/v0.0.3

The next version v0.1.0 will come with many new features including but not limited to:

• Acoustic models: multi-stream models (something similar to https://mtg.github.io/singing-synthesis-demos/), autoregressive F0 models
• Feature type: support for Mel-spectrogram
• Recipes: Chinese SVS Recipe for opencpop database #105
• Vocoder: harmonic-plus-noise uSFGAN
• Dev enhancement: dynamic mini-batch, distributed training Distributed data-parallel to speed up multi-GPU training #92

Please note that compatibility with the older versions is not guaranteed. It will likely to update ENUNU if you are an ENUNU user.

I'm going to start moving my local dev code to NNSVS. Note that It will take some time to complete.