LineIndex::offset used a position's character as is, without clamping it to the line. The language server protocol says a character past the end of a line means the end of that line, so these are valid requests that the server mishandled:

• Range formatting "ab\ncd" with the range (0, 100)-(1, 0) made the start offset (100) greater than the end (3) and hit the assertion in TextRange::new, which panics.
• An incremental change past the end of a line was applied at the wrong offset (silently corrupting the server's copy of the document) or panicked in String::replace_range when the offset was past the end of the text.
• A position between the two UTF-16 code units of a surrogate pair gave an offset in the middle of a character, which panics when slicing.
• A huge character value overflowed.

Now:

• A position past the end of a line is the end of that line, excluding the line ending.
• A position past the last line is the end of the text.
• A position in the middle of a surrogate pair is the start of that character.
• A range whose start is after its end is an error (logged, and range formatting returns no edits) instead of a panic.

LineIndex::offset no longer returns a Result because it can't fail anymore.