This is part of changing the annotator code to use a single
well-tested/debugged/optimized etc. implementation of Range <-> text offset
conversion in all places.