Implement KOReader's partial MD5 algorithm for document identification. This hash allows KOReader devices to match local files with server records for reading progress synchronization (KOSync).
Books may not have a path (e.g., physical books, metadata-only entries).
Updated path-dependent operations to handle None gracefully:
- get_file: raise ValueError if book has no path
- update_book: skip path relocation if no path exists
- remove_files: skip filesystem cleanup if no path exists
Also fixed _save_book_files return type and removed unused imports.
The create, update, and delete methods had incompatible signatures
resulting in typecheck errors. Renamed to create_book, update_book, and
delete_books.
The Darwin epub test cases used absolute paths. I hadn't notices until switching machines, whcih caused errors in the test harness. Changed to relative paths to be consistent with other test cases.