Estimating complex intervention effects as risk differences in cluster randomized trials