Decrease in gait variability while counting backward: a marker of “magnet effect”?