Clustering based pruning for statistical criticality computation under process variations