feat: add document discovery and search

[m1ngsama]

What changed

• add typed document parsing, metadata, semantic component extraction, section discovery, and ranked full-content search
• expose parseDoc, getDocument, listSections, and search as supported public APIs
• keep cached and in-flight requests bounded, with stale fallback behavior preserved
• add strict TypeScript, ESLint, Prettier, package-consumer checks, and a hardened trusted-publishing workflow
• prepare the package as @nbtca/docs@0.3.0 and keep the README focused on the public contract

Why

Consumers previously had to rebuild document parsing, routing, section discovery, and search around raw GitHub files. The expanded client centralizes those semantics so Prompt and other consumers can share one tested document model, including the latest components used by nbtca/documents.

Impact

This is a backward-compatible minor release. Existing listDir, listAll, and getFile consumers continue to work, while consumers can opt into richer discovery and search APIs. The published package remains renderer-agnostic and contains only runtime output, types, the README, license, and package manifest.

Validation

• npm run check
  • Prettier check passed
  • ESLint passed with zero warnings
  • strict TypeScript check passed
  • 71 tests passed across 3 test files
  • build and packed-package runtime/type consumer checks passed
  • npm audit --audit-level=moderate found 0 vulnerabilities
• npm publish --dry-run
  • passed the full prepublishOnly gate
  • confirmed @nbtca/docs@0.3.0, 13 published files, and a 10.1 kB tarball

This draft does not merge, tag, or publish the release.