Tag: rank by multiple columns in dplyr