On unbalanced group sizes in cluster randomized designs using balanced ranked set sampling