Self host models

Is there a way to host the models ourselves?
A way to replace the base urls for loading the voice models:
https://github.com/diffusion-studio/vits-web/blob/main/src/fixtures.ts#L6-L16

eg:
```
tts.setSources({
  HF_BASE: '...',
  ONNX_BASE: '...',
  WASM_BASE: '...' 
});
```

If not i might try to do a pull request for this is the authors agree it could be a usefull feature.