feat(skills): add auto-optimize workflow

[Qiong Wu (qiowu) (DingmaomaoBJTU)]

Summary

• check in the canonical auto-optimize Agent Skill from gim-home/ModelKitArtifacts@2bcf29d9150ab3af2909f3cc2a0b278d7ca2387d
• adapt the repository-layout contract and register skills/auto-optimize/tests in the existing CI matrix, following feat(skills): add model support contribution workflow #1415
• normalize imported Python to this repository's Ruff policy while preserving script interfaces and generated output
• close inherited fail-open validation paths for non-finite hotspot evidence and quoted/escaped absolute reproduction paths

Validation

• uv run pytest skills/auto-optimize/tests --tb=short --no-cov (309 passed)
• uv run ruff check skills/auto-optimize
• uv run ruff format --check skills/auto-optimize
• pre-commit hooks on all 13 target-adapted files
• all 7 bundled JSON documents parsed successfully
• all 29 pinned-source files are present; final diff is limited to the skill and CI registration

[Qiong Wu (qiowu) (DingmaomaoBJTU)]

Detailed readiness review: all 309 bundled tests passed locally, and all reported CI checks are green. However, the executable validation and workflow ordering gaps below prevent a ready-for-check-in verdict. Additional local reproductions confirmed acceptance of NaN final-report evidence and a generated replay wrapper returning success after an earlier native failure. No PR files were modified. The current tests provide useful script/contract coverage, but do not cover these end-to-end failure paths.