Optimizing RAM Usage When Calculating Maximum Value in Large Datasets with Dask and Pandas