feat: Add Markdown plain text projection

[Fuyeors]

Summary

This PR ports the TypeScript FFM Markdown parser to Rust and provides a search-oriented plain-text projection from the parsed AST. The projection is independent from Tantivy: the parser exposes text, while the search backend owns schema, tokenizer, field boosting, storage, and snippets.

to_plain_text()

The Rust crate now exports:

pub fn to_plain_text(nodes: &[AstNode]) -> String

The implementation performs one direct AST traversal. It keeps visible text, heading/paragraph/list/table/blockquote content, inline and fenced code, link labels, hard breaks, and FFM item titles/bodies. It omits Markdown presentation markers, link destinations, CSS class/style metadata, and structural thematic-break markers.

HTML and CSS are deliberately not parsed or sanitized by this API. When the parser represents them as text, they remain ordinary text and are returned unchanged. A canonical regression case covers a CSS <style> element. The function does not render HTML, strip tags, or depend on Tantivy.

Shared fixtures

fixtures/markdown.json and fixtures/ffm.json now include plain_text expectations for the standard Markdown and FFM cases. The shared cases cover headings, inline/fenced code, links, raw HTML/CSS text, tables, lists, hard breaks, FFM quote/accordion/chain/slide nodes, Unicode domains, and nesting behavior. TypeScript exposes a matching toPlainText reference implementation, and both language runners consume the same expected values.

Validation

• pnpm --filter @fuyeor/markdown-parser typecheck — passed
• pnpm --filter @fuyeor/markdown-parser test — 38 passed
• pnpm --filter @fuyeor/linkify test — 20 passed
• cargo fmt --all -- --check — passed
• cargo check --workspace --all-targets --all-features — passed
• cargo clippy --workspace --all-targets --all-features -- -D warnings — passed
• cargo test --workspace --all-targets --all-features — 2 thin runners passed

The existing CommonMark/GFM compatibility harness remains separate because it has its own filtered corpus and skip rules. The canonical cross-language suite is the source of truth for Rust/TypeScript plain-text parity.