Computes the I1 statistic for mean changes and generates
multiple candidate change locations using adaptive contrasts and local
interquartile scaling. Provides empirical I1 p-values for independent
Gaussian series of length 100 with full-window parameter 22, statistic-only
analysis for other settings, and weighted and unweighted segmentation
covering scores. DCI candidate detection is descriptive and does not
assign significance to selected segments. Based on the methodology of
Waychal, Laha and Sinha (2026)