Add a Hugging Face Spaces pre-requirements file to force a newer pip before installing app dependencies, and switch PyTorch model loading to download SuperAnimal weights into a writable local directory instead of the read-only site-packages model cache. The requirements update also pins Spaces to CPU-only PyTorch wheels to avoid pulling unnecessary CUDA packages.
Lock Gradio to the 6.29 release in both dependency manifests to keep installs consistent across environments, including Hugging Face Spaces. Also expand `.gitignore` for local virtualenv, tool caches, Gradio runtime data, downloaded prediction artifacts, and first-run DLC model downloads so generated files do not get committed.
Introduce a PyTorch inference path for SuperAnimal models, including backend selection in the UI, cached pose/detector runners, and JSON export of predictions. The legacy TensorFlow flow now maps model names separately from local directories, and MegaDetector device selection was corrected for yolov5 loading. The DeepLabCut dependency is also pinned to 3.0.2 to match the new backend integration.
