fix: handle character boundary in search mode

[roife]

fix #18918

I also used Clippy to scan and check usages of slicing in the project, and found that only here could cause a panic.

[ChayimFriedman2]

There is another slicing that looks like it may panic in line 305 (let prefix = &candidate[..query.len()];).

But do you know why it panics? We use match_indices(), it should return a correct index.

[Veykril]

But do you know why it panics? We use match_indices(), it should return a correct index.

We panic because we slice one byte after the character, so if the character is multi byte we slice into it instead of beyond it.

There is another slicing that looks like it may panic in line 305 (let prefix = &candidate[..query.len()];).

That one shouldn't cause issues, given the searcher in that mode does a prefix search and so the prefix should match fully (it would be a bug if not)

[ChayimFriedman2]

There is another slicing that looks like it may panic in line 305 (let prefix = &candidate[..query.len()];).

That one shouldn't cause issues, given the searcher in that mode does a prefix search and so the prefix should match fully (it would be a bug if not)

But it might not match.

[Veykril]

Then we shouldn't even reach that far, the fst automaton should only give us prefix matches, the only thing that can differ at that point is the ascii casing which will not cause issues here. (if we were doing general casing changes it might)