fix(extract): honor HTML base href for relative URLs

[Yoomitech]

Closes #42

Summary

Honor the first valid HTML <base href> element when resolving relative URLs during extraction.

Previously, relative references were always resolved against the fetched document URL. For pages using a base element such as:

<base href="/">

this could produce duplicated path components such as:

https://example.com/de/de/article.html

instead of:

https://example.com/de/article.html

Changes

• determine the effective URL base from the first valid <base href>
• use the effective base for extracted links and canonical URLs
• use the effective base for Markdown links and relative image URLs
• retain the fetched document URL as Trafilatura's OriginalURL
• fall back to the fetched document URL when no usable HTTP or HTTPS base is present
• add regression tests for metadata, canonical, Markdown link, and image resolution

Testing

• make fmt
• make lint
• make test
• real-world extraction test against a page using <base href="/">

All package tests and lint checks pass.