cl test incorrectly normalizes the observed likelihoods #78
Description
I just found a bug in the _poisson_likelihood_test function. For the CL-test, this function normalizes the observed likelihood score instead of comparing the score with the distribution of scores that would be expected if the model was correct, conditional on the number of observed earthquakes. Here, I am submitting a corrected version of the function that makes this little distinction for the CL-test, using an extra input variable CL=True.