icu_segmenter::LineSegmenter incorrectly applies rule LB8a

[tingerrr]

The following code using icu_segmenter = "1.3.2" returns 3 breaks:

• LineBreak::Unknown at 0
• LineBreak::ZWJ at 10
• LineBreak::Ideographic at 14

use icu_segmenter::LineSegmenter;

fn main() {
    assert_eq!(
        vec![0, 10, 14],
        LineSegmenter::new_auto()
            .segment_str("🏳️‍🌈")
            .collect::<Vec<_>>()
    );
}

According to the documentation, the segmenter should only return LB3 and LB7, not LB8a (LineBreak::ZWJ).

[sffc]

@makotokato @aethanyc @eggrobin

[eggrobin]

I see what’s going on, the way the state machine works effectively means that we apply LB9 before LB8a.