Hierarchical clustering has time complexity O(n³) for naive implementations. What makes this problematic for large datasets?