Update to 2026-09-17 pipeline snapshot; add weights, L20 assets and recording via Git LFS

Source: RGB-D -> Dyn-HaMR -> L20 retargeting -> FoundationPose -> reference repair -> SPIDER,
documented in docs/PIPELINE_LATEST.md and docs/SETUP_AND_WEIGHTS.md. Adds FoundationPose and
nvdiffrast upstream snapshots, requirements/pipeline_venv.txt and the FoundationPose weight
manifest/downloader.

Assets (Git LFS): weights/ (WiLoR detector, HandFlow denoiser, UniDepth-L), FoundationPose
checkpoints, HaMeR checkpoint, Dyn-HaMR HMP model and BMC constraints, L20 URDF/meshes, the
20260915_171525 D405 recording and the two box CADs. MANO models are not redistributed
(third_party/hamer/_DATA/data/mano/README.txt). Environments, caches and run outputs excluded.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

third_party/FoundationPose/environment.yml

@@ -0,0 +1,17 @@
+# Conda-only dependencies (compilers, CMake, Eigen, Boost, pybind11).
+# After creating the env, install PyTorch and GPU extensions as described in readme.md
+# (PyTorch index URL may need to match your CUDA driver), then:
+#   pip install -r requirements.txt
+#   bash build_all_conda.sh
+name: foundationpose
+channels:
+  - conda-forge
+dependencies:
+  - python=3.11
+  - pip
+  - cmake
+  - ninja
+  - eigen
+  - boost-cpp
+  - pybind11
+  - cxx-compiler