TaikiMiyagawa/Kaplan-Meier-Average-Run-Length ? reverse-engineered prompt

Build me a small research style notebook project that can evaluate quickest changepoint detectors using the KM ARL and KM ADD idea from the README. I want to be able to feed it sequences that may be different lengths, some with a changepoint and some without one, and get back average run length and average detection delay in a way that still uses the sequences that end before a detection happens. Please include a simple quick start notebook that shows the full workflow on example data, plus scripts that can generate the sample Gaussian and Poisson datasets and reproduce the evaluation results. If you need to, look up the current docs for the Python libraries used here so the setup actually runs cleanly. It would also be great if the code is organized so I can later plug in different changepoint detectors, including ones based on prebuilt chart statistics and ones where the statistic comes from the detection model itself.