* Added performance tests for hal::norm functions

* Added sum of absolute differences intrinsic

* norm implementation updated to use wide universal intrinsics

* improve and fix v_reduce_sad on VSX