Evaluating an Automatically Scorable, Open‐Ended Response Type for Measuring Mathematical Reasoning in Computer‐Adaptive Tests