Returns probability that a hypothesised segmentation *agrees* with a reference segmentation about whether points
a certain distance apart are in the same or different segment.
Given two probabilistic segmentations, returns the probability that points a certain distance apart are separated
by the same probability of segmentation.
Version which works on continuous segmentation probability distributions - arrays with one element per (usually
time) sample, with value = the probability that there is a topic shift at this element.