opart_gaussian          compute the optimal changepoint model for a
                        vector of real-valued data and a non-negative
                        real-valued penalty, given the square loss (to
                        minimize) / gaussian likelihood (to maximize)
opart_poisson           compute the optimal changepoint model for a
                        vector of real-valued data and a non-negative
                        real-valued penalty, given the poisson loss (to
                        minimize) / log likelihood (to maximize)
