* use codeunits/points instead of byte ranges when applicable * take into account different file formats when computing range and sending text (dos, unix, and mac supported) * add tests of incremental sync