minor Wolff optimization

We could exploit 1-exp(-x)<= x for x > 0. This would mean writing our own function that evaluates the exponential only as precise as required by the rolled random number. Of course it needs to be tested thoroughly