The minimization objective function in lmo.diagnostic.error_sensitivity is scalar (univariate). The EIF that this uses should be vectorized, which already is the case for those obtained from lmo.theoretical.l_{moment,ratio}_influence_from_cdf.

related to #362 and #330