Is the volume of a credal set a good measure for epistemic uncertainty?