Skipping unneeded partitions when joining dataframes in dask

Question

Consider the following scenario:

ddf_big = dd.read_parquet(\'...\', infer_divisions=True)  # large parquet with >500