Add story clustering to group duplicate stories across feeds

[samuelclay]

Summary

• New apps/clustering module that groups stories with matching or similar titles from different feeds using exact normalized title matching plus fuzzy Jaccard similarity on significant words
• Clusters stored in Redis (sCL: / zCL: keys, 14-day TTL) and displayed as always-expanded inline source list below the representative story in both river and single-feed views
• Celery task ComputeStoryClusters triggered after feed updates for feeds with premium subscribers, rate-limited to once per 6h per feed
• Briefing integration: shared normalize_title(), pre-computed cluster lookups in _find_duplicate_stories(), and cluster annotations in AI summary prompts

Test plan

• Open All Site Stories in river view, verify cluster sources appear inline below stories that have duplicates across feeds
• Click into a single feed that has clustered stories, verify clusters also appear there
• Verify cluster quality: all grouped stories should be about the same event from different feeds
• Check dark theme styling of cluster source rows
• Verify non-premium users do not see clusters
• Confirm Celery task runs after feed updates (docker logs newsblur_celery | grep Clustering)

---

Generated with Claude Code