Adding Confidence Intervals for Changepoint Detection

The central task in this project is to take the mentors’ existing confidence-set prototype and develop it into a maintainable package feature. The key challenge is to remove the current hardcoding around cost functions and replace it with a generic implementation that can support likelihood-based settings, including time series, while still fitting naturally into the existing changepoint workflow. I plan to do this by separating the core confidence-set logic from the algorithm- and cost-specific extraction steps, and then exposing the result through a user-facing interface with methods for summarizing, printing, and plotting the output. If successful, this would turn a research prototype into something that users of changepoint could actually apply in practice.