Alternatives to the Laplacian for Scalable Spectral Clustering with Group Fairness Constraints