fix: avoid mutable default arguments and use identity check for type comparison

[harshadkhetpal]

Summary

Three small correctness fixes flagged by ruff (B006 / E721):

1. src/tirith/core/core.py — start_policy_evaluation(var_paths: List[str] = [], inline_vars: List[str] = []) and start_policy_evaluation_from_dict(var_dict: Dict = {}) used mutable default arguments. Python evaluates defaults once at definition time, so the same list/dict object is shared across every call — a latent state-leak footgun, especially for var_dict, which is passed into get_policy_with_vars_replaced(). Changed to the Optional[...] = None sentinel pattern with normalization at the top of each function; behavior is unchanged for all existing callers.

2. src/tirith/core/evaluators/regex_match.py — type(evaluator_data) == str compares type objects with ==; identity (is str) is the intended semantics and is what newer linters (ruff E721) require.

Testing

python -m py_compile passes on both files and ruff check --select B006,E721 is clean after the change. No functional change for existing callers.

🤖 Generated with Claude Code

[sonarqubecloud]

❌ The last analysis has failed.

See analysis details on SonarQube Cloud