Fix extraction of reference links in Markdown

[mre]

Summary

Fixes #1657 by properly handling different types of Markdown reference links that were not being extracted by lychee:

• Reference links: [text][ref]
• Collapsed links: [text][]
• Shortcut links: [text]

Root Cause

The issue was in the markdown extraction logic where reference link destinations were processed using extract_raw_uri_from_plaintext(), which relies on the linkify crate. Linkify only recognizes URLs with schemes (http://, https://, etc.) and ignores relative file paths like "target.md".

Solution

1. Enable footnote support (ENABLE_FOOTNOTES) to properly differentiate footnotes from reference links in pulldown-cmark
2. Add explicit handling to skip footnote events (FootnoteReference, FootnoteDefinition) since they're not links to check
3. Create RawUri directly for reference links to handle relative file paths that linkify doesn't recognize

This approach leverages pulldown-cmark's built-in semantic distinction between footnotes and reference links rather than using heuristics.

Test Plan

• Added comprehensive test case covering all reference link types
• Verified footnotes are properly ignored (existing test still passes)
• All existing markdown extraction tests pass
• Manual testing confirms all four link types are now extracted correctly

Verification

Before:

[link1](target1.md) ✅ extracted
[link2][ref2] ❌ not extracted  
[link3][] ❌ not extracted
[link4] ❌ not extracted

After:

[link1](target1.md) ✅ extracted
[link2][ref2] ✅ extracted
[link3][] ✅ extracted  
[link4] ✅ extracted

The fix ensures lychee can now properly validate all types of Markdown reference links while maintaining backward compatibility.

[thomas-zahner]

That's great!