Release pure-Python ALDes 2.0.0 - #1

Summary

• replace the Python-to-MATLAB bridge with the pure-Python AutoOptLib 1.3.0 ALDes backend
• align generation and execution around the constrained paper-compatible fork/search/update grammar
• default to independent single-problem design without features, while retaining feature-conditioned EWC continual design
• automatically select CUDA/ROCm, Apple MPS, or CPU for PyTorch and use deterministic CPU multiprocessing for objective evaluation
• add a safe CLI, paper-subset runner, packaging, CI, citation metadata, and a 2.0.0 changelog
• remove bundled MATLAB sources, stale bytecode, obsolete loaders, and generated artifacts
• relocate all 335 historical reference-result files byte-for-byte and package them for portable comparisons

Validation

• ALDES_DEVICE=cpu ALDES_EVAL_WORKERS=2 python -W error -m pytest -q (12 passed)
• python -m ruff check .
• python -m ruff format --check .
• python -m build
• python -m twine check dist/*
• installed the 2.0.0 wheel in a clean Python 3.9 virtual environment outside the repository
• verified installed ALDes 2.0.0 resolves public AutoOptLib 1.3.0, both CLI entry points, and packaged 30-run paper reference data
• compared all 335 relocated reference files with the original Git blobs (0 mismatches)

Scope

ALDes 2.0.0 supports the paper's 23 PBO tasks. The original RIS beamforming and power-system restoration experiments are explicitly outside this release scope.