implement parser_fork.c

[Lakr233]

This will allow AI powered apps to render more efficiently. I'm new to cmark-gfm, so please review the code carefully. Thanks.

[Kyle-Ye]

Thank you for the contribution! I'd like to provide some context about this repository and your proposed feature.

Since this is a fork specifically maintained for the Swift toolchain and upstream swift-markdown package support, we generally welcome bug fixes but are more cautious about new feature implementations like parser_fork.c, as they may not align with our specific use case requirements.

A few questions to help evaluate this PR:

• Upstream consideration: Have you considered submitting this feature to the upstream repository (commonmark/cmark github/cmark-gfm) first? This would ensure broader community review and potentially benefit the entire cmark ecosystem.
• Swift integration: Do you have a corresponding PR planned for the swift-markdown package that would utilize this new API? Without concrete usage in the Swift ecosystem, it's uncertain whether this feature would be merged into our fork.
• Use case clarity: Could you elaborate on how this would specifically help "AI powered apps to render more efficiently"? Understanding the concrete use case would help us evaluate the necessity of this feature.

We appreciate your interest in contributing, but want to ensure any additions align with the Swift toolchain's specific needs and have clear integration paths with the broader Swift ecosystem.

cc @QuietMisdreavus

[Lakr233]

Well it sounds like pushing to upstream commonmark/cmark would be a better choice at this time. Let me do that first.

[QuietMisdreavus]

Also keep in mind that swift-cmark is itself a fork of github/cmark-gfm, which has different implementation concerns than commonmark/cmark (specifically the addition of extensions atop the CommonMark standard, like tables and strikethrough). I'm not opposed to adding this kind of API to swift-cmark, but i would also like to know the intended use case. There might still be some global state we inherited from GFM which would mess with this kind of desired deep copy functionality.