fix(convert): avoid unicode offset panic in attribute extraction

Author: chaliy

crates/fetchkit/src/convert.rs

@@ -414,9 +414,15 @@ pub fn html_to_text(html: &str) -> String {
 /// Extract attribute value from tag
 fn extract_attribute(tag: &str, attr: &str) -> Option<String> {
     let pattern = format!("{}=", attr);
-    let tag_lower = tag.to_lowercase();
-
-    if let Some(start) = tag_lower.find(&pattern) {
+    let start = tag
+        .char_indices()
+        .find_map(|(idx, _)| {
+            tag.get(idx..idx + pattern.len())
+                .filter(|candidate| candidate.eq_ignore_ascii_case(&pattern))
+                .map(|_| idx)
+        });
+
+    if let Some(start) = start {
         let rest = &tag[start + pattern.len()..];
         let rest = rest.trim_start();
 
@@ -1254,6 +1260,10 @@ mod tests {
             extract_attribute("div class=test", "class"),
             Some("test".to_string())
         );
+        assert_eq!(
+            extract_attribute("a title=\"İİ\" href=x", "href"),
+            Some("x".to_string())
+        );
     }
 
     #[test]