Computing heterogenous agent models when the distribution matters