feat(model): render and build documents outside the crate

[Idk121-blip]

What

Two additions to the public API, both Rust only:

• anydoc::document_to_markdown(&Document) -> String: the Markdown writer to_markdown_bytes already uses, callable on a document model directly.
• anydoc::model::GridBuilder: re-exported from model. new, next_row and finish get doc comments, which #![warn(missing_docs)] requires now that they are public.

The README's Rust section gets one line showing document_to_markdown.

Why

• A caller that edits a document from to_document, or builds one itself, cannot render it: the writer is reachable only through to_markdown_bytes, which parses first.
• A table built outside the crate cannot keep the grid invariant (every position exactly once, covered slots pointing at their origin) without copying GridBuilder. A parser for a format anydoc does not read is a frontend too, and the builder is how anydoc's own frontends get that guarantee.

How it was verified

• document_to_markdown_matches_the_direct_conversion: document_to_markdown(to_document(x)) equals to_markdown_bytes(x) for docx, epub, xlsx and csv fixtures.
• tables_built_outside_the_crate_keep_the_grid_invariant: builds a table with a column span from outside the crate, checks the covered slot points at its origin, and snapshots the rendered Markdown.
• cargo fmt --all --check, cargo clippy --workspace --all-targets --all-features -- -D warnings (Node, Python and wasm bindings included), cargo test --locked and cargo doc pass with no warnings.

Watch

GridBuilder was crate-private on purpose (pub(crate) use table::GridBuilder;, "frontends build grids; consumers read them off Table::grid"). This PR removes that line and its comment. If you'd rather keep the builder internal, document_to_markdown stands on its own and I can drop the GridBuilder part.

Follow-ups deliberately left out

• No binding changes: the Node, Python and wasm models are read-only copies, so there is no document to hand back to the writer.
• PDFs still have no document-model form, so document_to_markdown covers every format but PDF. That is the subject of PDF support via to_document / per-page extraction #62.

---

Summary by cubic

Lets callers render and build document models outside the crate. Previously the Markdown writer was reachable only through to_markdown_bytes, which parses first, so an edited or hand-built Document couldn't be rendered; document_to_markdown now renders a model the same way. GridBuilder is re-exported from model with its methods documented — it was crate-private, leaving external frontends no way to keep the grid invariant (every position exactly once, covered slots pointing at their origin) except by copying it. README and tests cover both additions.

^{Written for commit fcd14aa. Summary will update on new commits.}

[cubic-dev-ai]

No issues found across 6 files

Confidence score: 5/5

• Automated review surfaced no issues in the provided summaries.
• No files require special attention.

_{Re-trigger cubic}

[JonnyFox]

+1. We need parse → edit → render: today a Document modified after to_document can't be rendered without reimplementing the writer. document_to_markdown fixes that with no new surface.

GridBuilder public would be nice too, but if that's the contentious part, please land document_to_markdown on its own.

Thanks @Idk121-blip !!