feat: add Solidity smart contract security audit environment

[0xbyt4]

Summary

• New community environment that trains LLMs to detect security vulnerabilities in Solidity smart contracts
• Uses darkknight25/Smart_Contract_Vulnerability_Dataset (~2,000 labeled code snippets, MIT license)
• Multi-component reward function: vulnerability detection (0.25), category match with fuzzy matching (0.35), description quality via Jaccard similarity (0.25), format compliance (0.15)
• Follows existing patterns from sql_query_env and cybersecurity_sigma environments

Files

File	Purpose
solidity_audit_env.py	Main environment (BaseEnv subclass)
scoring.py	Reward function helpers
dataset_loader.py	HuggingFace dataset loading with robust JSONL parsing
test_scoring.py	46 unit tests for scoring logic
requirements.txt	Dependencies (datasets, pyyaml)
README.md	Documentation and usage instructions

Test plan

• 46/46 unit tests passing (pytest test_scoring.py -v)
• black, ruff, flake8 all clean
• Import check successful (from solidity_audit_env import SolidityAuditEnv)
• Process mode: dataset loads successfully (1998/2000 entries, 2 malformed rows skipped), API calls execute correctly (verified with Gemini API)