Read --datasets default value from DATASETS env variable

This allows running the pipeline of jobs with a DATASETS=WEO,WEOAGG environment variable, and process only those small datasets, avoiding waiting for all the SDMX datasets to be processed.