fix : fix broken prefix and suffix diffs generation inside addContext()

[michal-kurz]

Moved code here from this pr: #3

[michal-kurz]

Message by @dmsnell copied from original PR: @michal-kurz good catch on this. there's something else to consider here: considering that we're pulling in extra characters into the patch context, I guess there's still a possibility we might crash if the input document contains a series of surrogate halves. it might be that the existing fixup makes the same assumption, that if we shift one character we'll not find another lone low or high surrogate. just wanted to point it out.

Yes, I'm aware - but I believe this should always work as long as the input is valid. I also believe this also applies to your original solution for toDelta.

Fortunately for me, this is good enough for our use-case, as our app is not publicly accessible, and we are in direct contact with all our users - If our user somehow copies a naked surrogate pair (or other invalid text) into our text field, it's not a huge deal in our case. I don't expect this to ever happen, and if it does, then we will deal with it manually, or decide to fix this.