stop iterating through dataset after max_samples is reached and read language filter columns first (with license in the beginning now)