Handle zeros in probability distributions

We can do this by initially 'censoring' entries in a and b with value <= trim, and adapting c correspondingly. After the algorithm was applied, we then reintroduce the zeros.