Tags: kaya-go/katago-onnx

chore(dependencies): remove onnxconverter-common from pixi.lock

feat(upload): add --release-version flag to create HF tags after upload

feat(convert): add FP16 and UINT8 model exports with clean naming

- Rename export.py to convert.py for clarity
- Export three ONNX formats: FP32, FP16, and UINT8
- FP32 (.fp32.onnx): Recommended for browser/WASM
- FP16 (.fp16.onnx): For native apps with CoreML/CUDA/DirectML
- UINT8 (.uint8.onnx): For memory-constrained devices
- Add onnxconverter-common dependency for FP16 conversion
- Update HF_MODEL_CARD.md with new model format documentation
- Simplify upload to always delete existing files (delete_patterns='*')
- Add ruff lint config for import sorting
- Format all source files with ruff

docs: fix license reference in README (AGPL-3.0)