Fix cleanup pagination for seeded torrents

[CrusaderPistol]

Summary

Fixes #258.

The seeded torrent cleanup job previously limited its query to the first 100 eligible requests. Because each scheduled run could repeatedly query the same records, requests beyond that set could be starved indefinitely and never have their seeding requirements evaluated.

This change:

• Processes eligible requests in deterministic batches of 100
• Uses request ID progression instead of offset pagination
• Avoids skipping records when requests are hard-deleted during processing
• Tracks totalChecked across all processed batches
• Preserves existing shared-torrent protection and seeding behavior
• Adds a regression test covering more than 100 eligible requests
• Verifies that the second batch continues after the last processed request ID

Testing

• npx vitest run tests/processors/cleanup-seeded-torrents.processor.test.ts
  • 12 passed
• npm test
  • 209 test files passed
  • 2623 tests passed
  • 4 skipped
• npx tsc --noEmit
  • Passed
• docker compose -f docker-compose.local.yml build readmeabook
  • Build successful
• git diff --check
  • Passed