Testing multiple inequality hypotheses: A smoothed indicator approach