refactor(repo): finish cleanup and validation hardening

[LessUp]

Summary

• remove stale AI-process residue from docs and repo metadata
• fix CUDA validation issues uncovered by real release builds/tests
• harden package smoke and PyTorch integration validation

Validation

• cmake --preset release -DCMAKE_C_COMPILER=/usr/bin/gcc-12 -DCMAKE_CXX_COMPILER=/usr/bin/g++-12 -DCMAKE_CUDA_HOST_COMPILER=/usr/bin/g++-12
• cmake --build --preset release
• ctest --preset release --output-on-failure