Sample-Efficient Clustering and Conquer Procedures for Parallel Large-Scale Ranking and Selection