fix(utils): let profile URL regexps accept a trailing slash, query or fragment - #3344

URLMatcher.make_profile_url_regexp builds ^https?://(www\.|m\.)?<site>/<prefix>(?P<_maigret_username>...)$, so the username has to run all the way to the end of the link. Profile links found on pages (ids_links, which go through extract_ids_from_url) often don't:

• For a site with a regexCheck, https://github.com/soxoj/ or https://github.com/soxoj?tab=repositories doesn't match at all, so the username is lost.
• For a site without one, the default .+? swallows the rest: https://site.com/u/bob?ref=share gives the username bob?ref=share, which then gets searched on every site.

This allows an optional trailing slash and an optional ?query / #fragment after the username. Links that matched before still give the same username (a bare trailing slash was already stripped in detect_username), and characters the username regexp rejects still don't match.

Testing: new test_url_make_profile_url_regexp_allows_slash_query_and_fragment fails on main and passes with the change; the two tests that pin the exact pattern (test_url_make_profile_url_regexp, test_site_url_detector) are updated. Full pytest tests/: 554 passed, 5 skipped; the one failure, test_submit.py::test_check_features_manually_cloudflare, fails on main too.

This PR was prepared with AI assistance; I reviewed the change and ran the tests.