Replace medical image segmentation model with a new framework

Add ability to pre-download the segmentation model during build

Optimize model loading with dynamic quantization and device handling

Saved progress at the end of the loop

Improve model inference performance and reliability on hardware

Saved progress at the end of the loop

Improve AI segmentation by fixing errors and enhancing logging

Fix numpy version for Python 3.10 compatibility

Fix Dockerfile: replace libgl1-mesa-glx with libgl1

Add Dockerfile and README for Hugging Face Spaces deployment

Initial commit - InSono MVP clean