Standardized finish reasons

We've introduced finish_reason normalization to simplify the interpretation of responses:

• Normalized Finish Reasons: Models now return standardized finish_reason values, aligned with OpenAI-style conventions.

• Native Finish Reason: To retain model-specific details, the native_finish_reason is also returned.

Why This Matters

• Debugging and Understanding: Gain insight into a model's decision-making process.

• Cross-Model Consistency: Streamline development with unified finish reasons across reasoning-capable models.