Introducing the DeepSeek Harness Plugin: dsh-math-olympiad

`dsh-math-olympiad` is a mathematical reasoning plugin for DeepSeek Harness, specifically designed for competitive scenarios such as the IMO. It ensures the rigor and reliability of the problem-solving process through pure reasoning, adversarial verification, and calibrated confidence output. The plugin features LaTeX compilation and chain-of-thought isolation to prevent information leakage. Users should pay attention to pnpm version compatibility during installation. At runtime, it follows a "solve-verify-loop" process, allowing for honest abstention when a problem cannot be solved.