Optimized VPClus and VSPClus
by moving the first one to sort from stable_sort and by changing the comparison function to split between even and odd cases. On VPClus, this is a 20% speed improvement altogether. Probably much less on VSPClus as it was already using sort