Integrate anydoc document ingestion

[edison7009]

What changed

• Replace TierNote's lightweight DOCX/PPTX/XLSX/PDF text scraping with pinned anydoc 0.1.9 structured Markdown conversion.
• Add support for legacy Office, OpenDocument, RTF, EPUB, CSV, and additional OOXML variants while keeping conversion fully local.
• Preserve the native TierNote DOCX/PDF output generators; anydoc is used only for input ingestion.
• Add UTF-8-safe context truncation, document integration tests, file-type metadata, project handoff notes, and the upstream MIT notice.
• Remove the DeepSeek name from the Chinese empty-chat cache feature label.
• Resolve two Rust 1.93 Clippy style failures required by CI.

Why

The previous reader stripped XML tags or performed a minimal PDF scan, losing tables, lists, links, notes, and other structure. It also rejected legacy and OpenDocument formats. anydoc provides a tested local parser with fixed resource limits and substantially broader format coverage without sending user files to a hosted service.

User impact

The Agent can read more document formats and receives cleaner, structure-preserving Markdown. Existing document generation behavior remains available and unchanged. Scanned or image-only PDFs still require OCR.

Validation

• cargo clippy --manifest-path src-tauri/Cargo.toml --locked --all-targets -- -D warnings
• cargo test --manifest-path src-tauri/Cargo.toml --locked --lib — 187 passed
• npm run typecheck
• npm run library:check
• Rust formatting and git diff --check
• Upstream anydoc workspace tests — 208 unit tests plus robustness and snapshot suites passed

[edison7009]

Closed as superseded by 769cdb7. TierNote now uses its own local document-reading implementation, removes the AnyDoc dependency and notice, and completes the TierNote identifier migration without adopting the AnyDoc integration.