fix(html): preserve whitespace around anchor text

[IZHAR ALI JAFFAR (Izharalijaffar)]

HTML anchor conversion trims text with chomp(), but several return paths discard the removed whitespace. For example, <p>before<a> middle </a>after</p> currently becomes beforemiddleafter, losing word boundaries.

Preserve the original text when an anchor has no destination, contains only whitespace, or appears inside <pre>. Include the extracted prefix and suffix when emitting an autolink. Normal links and rejected JavaScript links retain their existing behavior.

Adds regression coverage through MarkItDown.convert_stream() for missing/empty hrefs, named anchors, whitespace-only anchors, preformatted text, and autolinks, plus controls for normal and rejected links.

Validation on Windows, Python 3.12.14, markdownify 1.2.3:

• Before the fix: 10 new regression cases failed; 8 cases passed.
• After the fix: all 18 HTML tests pass.
• Full core suite: 424 passed, 33 skipped, 3 failed. All three failures also reproduce with the original converter: test_file_uris and test_convert_case_insensitive_uri_schemes expect POSIX paths; test_doc_rlink attempts to write to C:\tmp\test_rlink.txt. Remote tests were skipped using the suite's GITHUB_ACTIONS setting; no live LLM calls were made.
• pre-commit run --all-files: passed.
• Wheel build: passed.
• Focused mypy check reports the same five existing options attribute errors on both original and changed converter; no new errors.

Prepared and tested with Codex assistance.

[IZHAR ALI JAFFAR (Izharalijaffar)]

@microsoft-github-policy-service agree