MS MARCO: Benchmarking Ranking Models in the Large-Data Regime