feat: download offline archives as files

[everpcpc]

Problem

Offline downloads for CBZ, CBR, and DIVINA-compatible EPUB books were still using page-by-page stream downloads. That made downloads slow and also diverged from Komga's server-side reading order, especially for DIVINA EPUB files whose manifest reading order may point at XHTML resources rather than image files.

Approach

Download supported archive-like formats through the book file endpoint and extract local reader images from the downloaded file. CBZ, CBR, and DIVINA EPUB extraction now use the server /pages metadata as the source of truth for image order and archive entry names. Background finalization writes sidecar metadata so a resumed app can finish extraction without relying only on in-memory state.

Scope

• Add Unrar.swift for CBR extraction.
• Route CBZ and CBR offline downloads through single-file archive extraction.
• Route DIVINA-compatible EPUB downloads through EPUB file extraction while keeping only image pages.
• Persist .archive-pages.json and .webpub-manifest.json sidecars for background download recovery.
• Mark WebPub, DIVINA, and generic JSON DTOs as nonisolated value models for Swift 6 Codable use.

Validation

• make format
• make build-macos
• make build-ios
• make build-tvos
• git diff --check