feat: add Readarr integration for managed downloads

[juchong]

Adds optional Readarr integration so ABR can hand off book requests to Readarr (or compatible forks like Bookshelf) for the full download-to-library pipeline. Closes #116

What this does

When Readarr is configured in Settings, new book requests are immediately sent to Readarr as a background task. Readarr handles searching, grabbing, downloading, importing, renaming, and organizing files. If Readarr is not configured, the existing Prowlarr auto-download flow is unchanged.

• New settings page (Settings > Readarr) for configuring the connection, quality profile, metadata profile, and root folder.
• Readarr search/add client that matches Audible books to Readarr metadata by title and author, adds new books, or triggers searches for existing ones.
• On successful Readarr add, the book is marked as downloaded and an ABS library scan is triggered.
• Fallback to Prowlarr in query_sources and download_book if the Readarr path fails.
• Pydantic models with extra="allow" for Readarr API responses, following the existing Prowlarr typing patterns (TypeAdapter, strict basedpyright).

What changed

• app/internal/readarr/ — new package: client.py, config.py, types.py
• app/routers/api/settings/readarr.py — API endpoints for Readarr settings
• app/routers/pages/settings/readarr.py — HTMX page routes
• templates/pages/Settings/Readarr.jinja — settings UI
• templates/layouts/SettingsLayout.jinja — added Readarr tab
• app/routers/api/requests.py — immediate Readarr handoff in create_request
• app/internal/query.py — Readarr-first path in query_sources for manual auto-download, increased client timeout to 300s
• README.md — documentation for setup and behavior

Checks

• basedpyright: 0 errors, 0 warnings, 0 notes
• ruff format --check app: all files formatted
• No new database models; no migrations needed