sched: minor optimizations in wake_affine and select_task_rq_fair