Comparing Methods for Bias Mitigation in Graph Neural Networks