fix(github): a multibyte char after < no longer panics the markdown sanitizer (#1076)

`starts_with_tag` sliced `rest[..name.len()]` by bytes, so an issue or
PR body with `<` followed by a multibyte char (e.g. `<日本`) panicked the
GitHub panel's render and quit the app. It now uses `str::get`, so a
non-boundary is just "not a tag". A regression test fails on the old
code and passes now.

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

crates/tty7-core/src/core/github/markdown.rs

@@ -256,8 +256,9 @@ fn starts_with_tag(s: &str, name: &str) -> bool {
     let Some(rest) = s.strip_prefix('<') else {
         return false;
     };
-    rest.len() > name.len()
-        && rest[..name.len()].eq_ignore_ascii_case(name)
+    // `get`, not a slice: `name.len()` can land inside a multibyte char.
+    rest.get(..name.len())
+        .is_some_and(|head| head.eq_ignore_ascii_case(name))
         && rest[name.len()..]
             .chars()
             .next()
@@ -609,6 +610,13 @@ mod tests {
         sanitize(src, "image")
     }
 
+    #[test]
+    fn a_multibyte_char_after_lt_is_prose() {
+        assert!(s("a <日本 b").contains("日本"));
+        assert!(!starts_with_tag("<日本", "img"));
+        assert!(starts_with_tag("<IMG src=x>", "img"));
+    }
+
     #[test]
     fn github_hosted_images_stay_images() {
         let pasted = "https://github.com/user-attachments/assets/352d14e0-d11c";