Parse EPUB structure (container.xml → OPF → spine → XHTML) to extract images in reading order. Zero new dependencies — reuses zip + regex crates with pre-compiled regexes and per-file index cache for performance. Falls back to CBZ-style image listing when spine contains no images. Includes DB migration, API/indexer/backoffice updates. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>