Mitigating Biases of Large Language Models in Stance Detection with Calibration