fix: keep escaped backslashes in angle-bracket link destinations

_parse_angle_link_href listed "\\" in its reject set, so any backslash
inside <...> made it return None and the whole link (or reference
definition) silently degraded to literal text. The bare-destination
branch in parse_link_href already consumes "\\" plus the next char, and
parse_link unescapes the result, so the escape machinery was already in
place and only this branch refused to feed it.

Consume the escape as a unit like the sibling branch does; an unescaped
"<", line ending, or NUL still terminates as before. Matches CommonMark,
which allows escaped "<"/">" inside an angle-bracket destination.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Author: chuenchen309

src/mistune/helpers.py

@@ -242,7 +242,10 @@ def _parse_angle_link_href(src: str, pos: int) -> Union[Tuple[str, int], Tuple[N
         c = src[pos]
         if c == ">":
             return src[start:pos], pos + 1
-        if c in "<\\\n\r\x00":
+        if c == "\\":
+            pos = min(pos + 2, len(src))
+            continue
+        if c in "<\n\r\x00":
             return None, None
         pos += 1
     return None, None

tests/test_misc.py

@@ -134,6 +134,14 @@ def test_link_bracket_cache_cases(self):
         for text, html in cases.items():
             self.assertEqual(mistune.html(text).strip(), html)
 
+    def test_angle_link_escaped_backslash(self):
+        cases = {
+            r"[link](<foo\*bar>)": '<p><a href="foo*bar">link</a></p>',
+            "[x]: <a\\*b>\n\n[x]": '<p><a href="a*b">x</a></p>',
+        }
+        for text, html in cases.items():
+            self.assertEqual(mistune.html(text).strip(), html, text)
+
     def test_allow_harmful_protocols(self):
         renderer = mistune.HTMLRenderer(allow_harmful_protocols=True)
         md = mistune.Markdown(renderer)