Move RSS and daily drains into the persistent worker

[BoraGkc]

Why

rss-cron.yml and full-cron.yml ran in this public repo and curled Cloud production. When the Hetzner host lost network on 2026-09-25, they started failing and emailing. Production scheduling belongs with production.

What

• background-worker.ts: a non-blocking periodic scheduler inside the existing worker loop.
  • feeds every 6 h (fixed cadence unchanged), using RSS_CRON_MAX_FEEDS / RSS_CRON_MAX_POSTS_PER_FEED through readCronDrainConfig.
  • daily: Google indexing, Search Console sync, and expired operation_events purge (previously only the all-task drain purged them).
  • Both run once at worker start; runScheduler due checks and feed claims keep that safe. An unfinished run is never started twice, and a failure waits for the next interval.
• Deleted rss-cron.yml, full-cron.yml, and campaign-cron.yml.
• Updated architecture.md, operations.md, and rss-scheduler.md.

Self-hosted Compose, Dokploy, and Railway schedulers are unchanged. The Cloudflare fallback Worker is unchanged.

Verification

• npm run test:server: all 59 self-tests pass, including new periodic-scheduler cases
• Server tsc --noEmit is clean; git diff --check is clean

Deploy note

RSS has no scheduler between merging this and the Cloud worker image rolling out. Deploy promptly after the Cloud sync. Then remove the now-unused BLOGFACTORY_CRON_SECRET repo secret.

🤖 Generated with Claude Code