Optionally keep original title headers for main content extraction accuracy

[mcPear]

Summary

Reader extraction currently rewrites all in-article h1 elements to h2 so the article title can remain the sole top-level heading in the reader UI. Moreover, removes the first similar heading spotted after the title. These normalizations improve classic “reader mode” presentation but weaken the semantic outline of the page: crawlers, SEO tooling, and systems that infer structure from HTML (including retrieval and “reverse engineering” of how a page is organized) rely on stable heading levels that match the publisher’s markup.

This change preserves the original heading tag names and levels in the extracted content wherever we are not explicitly removing noise, so the serialized article HTML stays closer to the source document’s hierarchy. All that is gated behind an option.

What changes (high level)

• Stop blanket h1 → h2 replacement in article content
• Stop duplicate-title header removal
• Add unit tests

[gijsk]

Thanks for submitting a PR.

I'm confused about this being labeled a "chore". And I think that the relabeling of h1 to h2 and removing headers that match the titles should be split into different options. The h1-to-h2 replacement would fix #863 . I would be tempted to make not swapping it the default, though it would potentially break existing consumers - I can fix Firefox but not sure about others.

[gijsk]

Apologies for the very unreasonable delay here.

I'm fairly confused at this point. The updated patch here added a lot more code which is a bit hard to read, and also doesn't seem consistent with what I asked in my previous review. Instead, it seems to change the approach used to keep the headers. Why is that?

[mcPear]

Sorry, I had to adjust the logic to fix corner cases we found in our app. I wasn't aware you will see it as kinda response to this conversation. I'm closing it as for now.