feat(video): preserve audio track and fix video frame assembly

feat(video): add /pipeline_video endpoint with frame-by-frame annotation and H264 output

Add HF Spaces README with Docker sdk configuration

fix(detect): raise confidence threshold 0.25→0.45 and iou 0.45 to cut false positives

fix: use /tmp/models for writable model cache, hardcode HF repo

fix: rename env vars to alphanumeric-only for HF Spaces secret naming rules

fix(startup): always load pretrained yolov8n fallback regardless of file existence

fix(docker): replace libgl1-mesa-glx with libgl1 for Debian trixie

deploy: support all models (YOLO, FFA-Net, Zero-DCE++, Restormer, buffalo_l)

deploy: add HF Hub model download on startup

fix

add deploy/ folder — HuggingFace Spaces inference API