Which method is better and more efficient than the other, to choose a pivot randomly or to choose it by selecting three indexes (index 0, index n/2 and index n) and taking the middlemost number between the three?