Automatic Data Curation for Self-Supervised Learning: A Clustering-Based Approach