test: Optimize test_batch_inference to actually perform concurrent stress testing using asyncio.gather

chore: Sync all latest updates

fix: Add httpx retries to prevent RemoteProtocolError from flaky HCMC server

fix: Add opencv-python-headless to requirements.txt for heatmap generation

feat: Add OpenCV-based heatmap_base64 generation endpoint

feat: implement image fallback cache to handle intermittent camera API disconnects gracefully

feat: release final optimized ONNX 320x320 pipeline

feat: add batch inference testing script and update documentation

fix: resolve ModuleNotFoundError for models import

perf: optimize cpu inference speed

fix: resolve CORS and swagger docs issues

chore: add .gitignore + update ZIP model checkpoint

perf: add 120s TTL cache for batch_predict to prevent CPU overload and reduce batch max to 15

feat: upgrade ZIP prediction model, add batch_predict API, caching and concurrency semaphores

Fix multi-class output: read per-class density maps (ch0=car, ch1=motorbike) instead of heuristic 70/30 split

Fix ZIP_PROJECT_ROOT path resolution for Docker structure

Add required HF Space README config

Fix HF Space dependencies

Initial commit for HuggingFace Space backend