Add a new `--keep-filenames` boolean flag (no shorthand, default false) on
both `optimize` and `watch` commands. When enabled, page entries in the
output CBZ preserve their original base filenames instead of being
renumbered to `%04d` sequential indices. Extension is swapped on format
conversion (e.g. `001.png` → `001.webp`), split pages get `-NN` suffixes
on the original stem, and duplicate base names fall back to
`<stem>_<index>.<ext>`. Watch mode supports the flag via the
`keep-filenames` viper config key.
Design: data-driven `OriginalName` field on `manga.PageFile`, set at
extraction in `cbz_loader.ExtractChapter`. The `Converter` interface is
unchanged. Flag off → behavior identical to before.
Closes#216
BREAKING CHANGE: Complete refactoring of the conversion pipeline.
- Replace in-memory Page/PageContainer with disk-based PageFile struct
- Extract archives to temp directory instead of loading into memory
- Use cwebp InputFile/OutputFile for zero-copy file-to-file conversion
- Use cwebp -crop for splitting (no Go-side image decode needed)
- Check if already converted before extracting (fast pre-check)
- Remove oliamb/cutter dependency (replaced by cwebp -crop)
- Remove page_container.go (no longer needed)
- Add comprehensive tests with improved coverage
Co-authored-by: Belphemur <197810+Belphemur@users.noreply.github.com>
Antoine Aflalo