Unicode Matching Indices are potentially incorrect

[MinusGix]

fn main() {
    let matcher = SkimMatcherV2::default();

    let text = " üäö ";
    // [(0, ' '), (1, 'ü'), (3, 'ä'), (5, 'ö'), (7, ' ')]
    println!("{:?}", text.char_indices().collect::<Vec<_>>());
    println!("{:?}", matcher.fuzzy_indices(text, "ü")); // -> 1 (good)
    println!("{:?}", matcher.fuzzy_indices(text, "ä")); // -> 2 (bad)
    println!("{:?}", matcher.fuzzy_indices(text, "ö")); // -> 3 (bad)

    let text = "2  üäö ";
    // [(0, '2'), (1, ' '), (2, ' '), (3, 'ü'), (5, 'ä'), (7, 'ö'), (9, ' ')]
    println!("{:?}", text.char_indices().collect::<Vec<_>>());
    println!("{:?}", matcher.fuzzy_indices(text, "ü")); // -> 3 (good)
    println!("{:?}", matcher.fuzzy_indices(text, "ä")); // -> 4 (bad)
    println!("{:?}", matcher.fuzzy_indices(text, "ö")); // -> 5 (bad)
}

I would have expected the indices to match the 'indices' into a string, but it seems to be matching the indices into the chars()/chars_indices() iterator.
Is this intended behavior (which should probably be documented?), or is it a case of missing a + ch.len_utf8() in order to properly increment the index into the string?

[RobWalt]

Maybe there should be an extra option for that behavior? I could imagine that people find the current style of handling things usefull for other use cases.