Add Markpub content parser

[pfefferle]

Proposed changes:

• Add Markpub parser (at.markpub.markdown) that walks Gutenberg blocks via parse_blocks() and converts each to GFM markdown.
• Register Markpub as the default content parser via the atmosphere_content_parser filter.
• Add tests for all supported block types and filter integration.

Stacked on #8. Merge that first.

Other information:

• Have you written new tests for your changes, if applicable?

Testing instructions:

• Verify ?atproto preview on a published post returns valid JSON with a content field containing at.markpub.markdown data.
• Verify block content converts to markdown correctly (headings, links, images, lists, code blocks).
• Verify classic editor content falls back to inline HTML-to-markdown conversion.
• Verify atmosphere_content_parser filter with __return_null disables the content field.
• Run npm run env-test — all tests pass.

Changelog entry

• Automatically create a changelog entry from the details below.

Changelog Entry Details

Significance

• Patch
• Minor
• Major

Type

• Added - for new features
• Changed - for changes in existing functionality
• Deprecated - for soon-to-be removed features
• Removed - for now removed features
• Fixed - for any bug fixes
• Security - in case of vulnerabilities

Message

Add rich content support for standard.site documents using the Markpub format.

[kraftbj]

Thanks for this — the parser structure is clean and the namespaced-prefix conventions look good throughout. Left a few inline notes on some edge cases I spotted while reading; none of them block the overall approach.

[pfefferle]

Thanks for the review! All five points addressed in 8991043:

L36 phpcs:ignore rule + $post comment — both Generic.CodeAnalysis.UnusedFunctionParameter and VariableAnalysis.CodeAnalysis.VariableAnalysis.UnusedVariable actually fire on this parameter, so the ignore now lists both. Added a docblock note explaining that $post is required by the Content_Parser contract even though Markpub doesn't use it.

L65 advertised table extension — dropped from the extensions array. No core/table handler yet; the advertised capability should match what's actually emitted. Can re-add when a table transform lands.

L66 empty-markdown null sentinel — widened Content_Parser::parse return type to ?array. Markpub returns null when the trimmed markdown is empty; Document::get_content bails early on null and skips the atmosphere_document_content filter (keeping callbacks' array $content signature honest). New tests cover a post of only core/spacer blocks, literal empty post content, and the Document-side null-handling path (asserting the filter is not invoked).

L164 figcaption sibling-bleed — took your suggested order swap: preg_replace the surrounding figcaption markers first, then wp_strip_all_tags. Added a regression test with a trailing <p> inside the same <figure>; without the fix it fails with "Real captionShould not appear in caption" concatenated.

L367 preg_replace PCRE guard — added safe_replace / safe_replace_callback helpers that fall back to the unchanged subject and wp_trigger_error when preg_replace returns null. Applied to all inline conversions and the figcaption chain. Didn't TDD this one directly — reliably tripping PCRE limits in a unit test is awkward — but the new call sites are uniform and the helper is small enough to read at a glance.

[pfefferle]

@kraftbj we should maybe find a more generic way, to also transform into the other standard.site content formats.

[kraftbj]

Agreed — the single-parser filter and one-NSID-per-implementation shape will get awkward as soon as a second standard.site format lands. Filed #45 to nail down the interface before that happens (registry vs per-NSID filter vs canonical-IR projection — open for discussion).

For this PR, let's move forward with the current shape and iterate. Markpub is the only parser today, so the registry refactor doesn't block anything user-facing, and #45 keeps the design conversation in its own thread.