Let Rust own shortcode expansion: fix escaped shortcodes crashing pandoc, and expand them in footnote definitions

[gordonwoodhull]

Summary

Escaped shortcodes such as `{{{< brand >}}}` crashed pandoc with exit 83 whenever the page was rendered through the pandoc path. Rust unescapes {{{< x >}}} once, and the vendored Lua shortcodes pass then expanded the result as a live shortcode. For handlers that Rust doesn't know (brand), the Lua side hit a nil brandCommand at shortcodes-handlers.lua:111.

The fix makes Rust the only place shortcodes are expanded. Two commits:

1. Expand shortcodes in footnote definitions and lst-cap (bd-xjg7vl6c). shortcode_resolve skipped NoteDefinitionPara, NoteDefinitionFencedBlock and the lst-cap attribute of listing code blocks, which the Lua pass had been covering. Quarto 1 expands all three, so Rust now does too. This has to land first. Without it, turning Lua off would leave [^1]: Note {{< meta author >}}. unexpanded.
2. Disable the vendored Lua shortcodes pass (bd-2uva9urq). The pre-shortcodes-filter entries in main.lua and crossref/crossref.lua now require q2_lua_shortcodes_disabled, a flag that is never set, so the runner skips them. This is a flag-gate rather than a removal. shortcodes_filter() must still be called because foldcode.lua's process_shortcodes uses _shortcodes_filter. Both edits carry a QUARTO2-PATCH comment, and the README's "Ours vs. pinned" section documents them.

Tests

• shortcode_all_contexts: a parameterized test over every context where a shortcode can appear. Contexts that Quarto 1 also leaves literal are pinned as literal.
• shortcode_pandoc_escapes: renders to native and checks that escaped brand shortcodes stay literal in a code span, a code span with arguments and a fenced block. A control confirms that a live {{< brand >}} still only warns. Three of the four cases failed with exit 83 before the flag-gate.
• pandoc_filters::test_shim_group_shape_and_patch_markers pinned exactly two QUARTO2-PATCH markers in main.lua. It now ignores the unrelated shortcodes marker.

Not in this PR

• Brand support. A live {{< brand >}} still doesn't render. That stays with bd-qnylgu69.
• pandoc-wasm. pandoc-request.golden.json stores a hash of the embedded filter tree (share_tree_version), so editing the Lua changes it. That file only exists on feature/pandoc-wasm. Whichever branch lands second needs one Q2_REGENERATE_GOLDEN=1 regeneration. The pandoc recordings are self-contained and don't need re-recording.

Test plan

• cargo nextest run --workspace (running locally; result will be posted as a comment)
• CI green

[posit-snyk-bot]

✅ Snyk checks have passed. No issues have been found so far.

Status	Scan Engine	Critical	High	Medium	Low	Total (0)
✅	Open Source Security	0	0	0	0	0 issues
✅	Licenses	0	0	0	0	0 issues

💻 Catch issues earlier using the plugins for VS Code, JetBrains IDEs, Visual Studio, and Eclipse.