[Hilda-Azimi]

Adds examples/dlc-lightning-pose-env/: a self-contained, uv-locked Python 3.11
environment in which DeepLabCut 3.0.2 and Lightning Pose 2.4.2 install and run
together (156 pinned packages; torch 2.14.1, lightning 2.5.6, numpy 2.4.6, a
single OpenCV build). Shared as an example — it does not change poseinterface's
own dependencies.

Install: uv sync --frozen --no-install-project, then uv run python smoke_test.py.
Verified on Linux + NVIDIA T4 (both tools train, evaluate and run video inference)
and on Apple Silicon.

Notes

• The [tool.uv] override-dependencies block resolves three conflicting OpenCV
  distributions pulled in by the two tools; opencv-contrib-python-headless is
  the single build used.
• NVIDIA DALI (Lightning Pose's video reader) is Linux x86_64 only; LP falls back
  to OpenCV elsewhere.
• uv lock must be run on Linux (DALI's PyPI stub must execute).
• Pinned to Python 3.11, while poseinterface targets >=3.12. Happy to re-lock for
  3.12 if you'd prefer this integrated as a dependency group rather than an example.