This branch adds
-a LogLikelihoodSum wrapper that takes multiple LogLikelihood objects and can be passed to optimisers, inference methods to compute combined fits.
-by setting the flag allowed_negative for a source, a rate multiplier for this source will be allowed to be negative. Checks for positive rates are bypassed for this source, while the total rate must still be positive. Setting the outlier_likelihood forces the "local" pdf to be positive at each data-point. So far, this has been written only for unbinned likelihoods.
This branch adds -a LogLikelihoodSum wrapper that takes multiple LogLikelihood objects and can be passed to optimisers, inference methods to compute combined fits. -by setting the flag allowed_negative for a source, a rate multiplier for this source will be allowed to be negative. Checks for positive rates are bypassed for this source, while the total rate must still be positive. Setting the outlier_likelihood forces the "local" pdf to be positive at each data-point. So far, this has been written only for unbinned likelihoods.