Diffusion-Based Clustering of High-Dimensional Datasets