Calibrated Stackelberg Games: Learning Optimal Commitments Against Calibrated Agents