feat(models): add max_completion_tokens to model definitions and responses