New SOTA music separation model

[lucianodato]

I´ve been messing around with UVR recently and discovered that there´s this new model for music separation that gives better results that hybrid demucs. I wanted to highlight it to check if there´s interest to add it to the project.

https://github.com/starrytong/SCNet

Please feel free to close the issue if this is either not possible or not something of the interest of the project currently.

Thanks for all your efforts!