fix(sitemap): Prevent sitemap parser from leaking metadata between <url> entries

[anxkhn]

Description

The XML sitemap parser leaks per-entry metadata from a malformed <url> block into the next entry.

• In _XMLSaxSitemapHandler.endElement (src/crawlee/_utils/sitemap.py), the emit and the reset of self._current_url were both gated behind if name == 'url' and 'loc' in self._current_url:.
• _current_url accumulates lastmod / priority / changefreq as those child tags close. When a <url> block has metadata children but no (or an empty) <loc>, the closing </url> did not reset _current_url, so its stale metadata carried over and attached to the next parsed <url>.
• Fix: reset _current_url on every closing </url>, while still only emitting an item when a loc is present. This discards a locless block's metadata instead of leaking it, and leaves well-formed sitemaps unchanged.

-        if name == 'url' and 'loc' in self._current_url:
-            self.items.append({'type': 'url', **self._current_url})
+        if name == 'url':
+            if 'loc' in self._current_url:
+                self.items.append({'type': 'url', **self._current_url})
             self._current_url = {}

Issues

• No existing GitHub issue. This is a self-identified correctness bug found while reviewing the sitemap parser. Happy to open a tracking issue with the repro first if you'd prefer that.

Testing

• Added test_xml_handler_discards_metadata_from_url_without_loc to tests/unit/_utils/test_sitemap.py. It drives the real ExpatParser + _XMLSaxSitemapHandler path (the same path _XmlSitemapParser uses in production), feeds a locless metadata <url> followed by a real <url><loc>...</loc></url>, and asserts the output is a single clean item with no leaked lastmod / priority / changefreq.
• Verified fail-first: the new test fails on master (the metadata leaks) and passes with the fix.
• uv run pytest tests/unit/_utils/test_sitemap.py tests/unit/request_loaders/test_sitemap_request_loader.py -> 101 passed. The request-loader suite exercises the same parser path.
• uv run poe lint (ruff check + format) and uv run poe type-check (ty) both pass on the changed files.

Checklist

• CI passed

[Mantisus]

Hey, @anxkhn. Thank you for your contribution.

Only one moment.

[Mantisus]

LGTM

[vdusek]

LGTM