MultiByte characters lead to offset errors

[alemidev]

Inserting an emoji works correctly and sends it to remotes, but deleting it gets sent incorrectly and leads nvim out of sync. Also if we write an emoji and someone else deletes it, we get broken chars:

[alemidev]

Investigating this more in-depth, codemp indexes buffers by char (and so do most other editors), but nvim does so by byte, leading to out-of-sync insertions after multi-bytes characters in buffers.

codemp-nvim-demo-multibyte-desync.mp4

There is a charidx() builtin but requires a whole string in input