Added changes related to deployment on hugging face space

- Dockerfile: setting HF_HOME and TRANSFORMERS_CACHE to /tmp, we avoid permission issues
- README.md: include required metadata block at the top of the file

Dockerfile

@@ -27,4 +27,4 @@ COPY fe/ ./frontend/
 EXPOSE 8000 7860
 
 # Run both backend and frontend in parallel
-CMD ["sh", "-c", "python run.py & python -m frontend.gradio_app"]
+CMD ["sh", "-c", "python run.py & python -m frontend.gradio_app"]