Release Notes: v4.0.2

🚨 Critical Updates

These changes are essential for database stability and compatibility with Calibre 9.

Calibre 9 Metadata Compatibility: Fixed critical UI crashes (500 errors) caused by Calibre 9's removal of books.isbn, books.flags, and books.lccn.
- The system now dynamically detects schema versions at startup.
- For Calibre 9+, ISBNs are now correctly pulled from the identifiers table.
- Maintained full backward compatibility for users on older Calibre versions.
SQLAlchemy 2.x Migration Fix: Resolved InvalidRequestError during database migrations by utilizing engine.begin(). This prevents nested transaction failures when PRAGMA triggers autobegin.
Initialization Fix: Fixed a regression causing AttributeError during login. The system now ensures minimal config fields (like config_books_per_page) are loaded even during background initialization.

Route Reorganization: * Renamed the versioning page to /package-versions to reduce ambiguity.
- Redirected legacy /stats path to /cwa-stats-show.
Docker Updates:
- Updated Dockerfile to Calibre 9.0.0.
- Unified Calibre/Kepubify build arguments to ensure the UI accurately reports installed versions.
Fork Modernization: Removed "Stock Calibre-Web" version display from admin stats as it is no longer applicable to this fork.

OAuth Link Loop: Fixed a "UNIQUE constraint failed" error when linking an OAuth provider to an existing account.
Hardcover Auto-Fetch: Fixed a bug where auto-fetching would ignore cancel commands and block the task queue.
ARM Builds: Fixed network configuration issues in the ARM build workflows.

Automation: Updated translation workflows to automatically commit .po and .pot files.
Build Flow: Translation updates now trigger dev builds automatically to keep the main branch images up to date with the latest localization.
Issue #960 Resolution: Addressed a major translation regression introduced in 4.0.0/4.0.1 that caused broken UI strings across multiple languages.
Babel Execution Fix: Switched translation scripts to use python -m babel invocation. This bypasses issues with broken virtual environment shebangs that were previously causing automation failures in the CI/CD pipeline.

Image Optimization: Updated the dev workflow to skip redundant "main push" builds when a translation build is already triggered, preventing duplicate Docker images and saving runner resources.
Docker Cleanup: Dropped the unused CW base label and arguments from the Dockerfile to streamline the multi-stage build process.
Migration Testing: Added a specific "smoke test" for the database transaction pattern to ensure future schema changes don't re-introduce the SQLAlchemy InvalidRequestError.