Unquantized model files aren't cleaned up after conversion

[alexrozanski]

When importing .pth models we convert them to .ggml files first then quantize them; however, these un-quantized models aren't cleaned up so sit in the internal models directory, even though they are not needed.

[alexrozanski]

This should be an easy fix:

• Ensure they're cleaned up
• Run a function on app launch to clean up any unquantized models in the internal models directory, to mitigate this issue for existing users