powerpc: Optimize memrchr for power8

Vectorized loops are used for sizes greater than 32B to improve
performance over power7 optimization.  This shows as an average
of 25% improvement depending on the position of search
character.  The performance is same for shorter strings.