Avoid slicing text inside multi-byte characters

Right-to-left line reordering walked each atom's byte range and sliced the
line at every byte, panicking when a neutral multi-byte character such as a
guillemet preceded the first strong character. Walk characters instead.

FIND/FINDALL located the REPLACE keyword in a Unicode-uppercased copy and
sliced the original with that offset; uppercasing can change byte length
(dotless i, ligatures). Uppercase ASCII only, which keeps offsets aligned.
FILLET and CHAMFER inline R/D shorthands now test the original text before
slicing past the first character.

Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>

src/entities/text_support.rs

@@ -798,9 +798,11 @@ pub fn reorder_line_atoms(atoms: Vec<LayoutAtom>, is_rtl: bool) -> Vec<LayoutAto
         .iter()
         .map(|r| {
             let mut lvl = None;
-            for i in r.clone() {
+            // `bidi.levels` is indexed by byte, so walk characters rather than
+            // bytes to avoid slicing inside a multi-byte character.
+            for (offset, c) in line_text[r.clone()].char_indices() {
+                let i = r.start + offset;
                 if i < bidi.levels.len() {
-                    let c = line_text[i..].chars().next().unwrap_or(' ');
                     match unicode_bidi::bidi_class(c) {
                         unicode_bidi::BidiClass::L
                         | unicode_bidi::BidiClass::R
@@ -2966,6 +2968,12 @@ mod v_anchor_tests {
             reordered_b,
             vec!["2026 ", "is ", "here!", "שנה ", "שלום ", "مرحبا! ", "ہے۔ ", "اردو ", "یہ "]
         );
+
+        // Case 6: a word starting with multi-byte neutral punctuation must
+        // not be sliced inside that character.
+        let words6 = ["שלום", " ", "«x"];
+        let reordered6 = reorder_atoms(&words6, false);
+        assert_eq!(reordered6, vec!["שלום", " ", "«x"]);
     }
 
     #[test]