Update to darkseid 9.0.0

[bpepple]

Summary

Updates darkseid from 8.4.1 to 9.0.0 and makes page hashing handle unreadable pages explicitly.

darkseid 9.0.0 API change

darkseid#253 limits uncompressed archive reads to guard against decompression bombs:

• Archiver.read_file() has a new max_size parameter (default 256 MiB per file). Oversized files raise the new ArchiverFileTooLargeError, a subclass of ArchiverReadError.
• ComicInfo.xml and MetronInfo.xml are read with a tighter 10 MiB limit.

This API change only breaks custom Archiver subclasses. metron-tagger has none and never calls read_file() directly, and the existing ArchiverReadError handler in filesorter.py also catches the new error.

Changes

• Raise the darkseid requirement to >=9.0.0 and update uv.lock.
• Comic.get_page() returns None when a page can't be read, which now includes pages over the size limit. Before, a None page only worked because an empty BytesIO happened to raise an image error that was already caught. The code now checks for None explicitly:
  • Duplicates._calculate_image_hash() takes bytes | None and returns None for unreadable pages.
  • CoverHashMatcher.get_comic_cover_hash_cached() reports "Unable to get cover hash." when the cover is missing.
• Add a test for None page data in duplicate hashing.

Behavior notes

• Metadata files over 10 MiB now read as empty metadata.
• Comic.validate_metadata() now returns UNKNOWN on read errors instead of raising AttributeError. With --validate, such files are reported as not valid, and removed if the user asked to remove invalid metadata.
• darkseid 9.0.0 drops Python 3.10. metron-tagger already requires 3.11 or later, so nothing changes here.