v0.2.0

Highlights

• Added per-run token usage and cost tracking in simulation outputs (meta.json).
• Added schema-driven categorical null phrasing (null_options / null_phrase) for persona rendering.
• Added dependency auto-inference during constraint binding and related validator coverage.
• Removed checked-in study artifacts from the core package repo.

Validation

• ruff check .
• ruff format --check .
• pytest -q (637 passed)