shared hashes: separate background job with visible state

- compute duplicate-hash statistics in their own job (startup, hourly,
  on demand) so they no longer block the table stats
- single-pass query over DISTINCT (hash, addon_id), 15 min timeout
- record errors and panics; never stay stuck in a computing state
- whitelist page polls while computing, shows errors and a Recompute
  button; tab is linkable via ?tab=shared

admin/system.go

@@ -199,7 +199,8 @@ func maintenance(c *gin.Context) {
 	}
 	switch body.Action {
 	case "refresh-stats":
-		requestHeavyRefresh(true)
+		requestHeavyRefresh()
+		requestDupRefresh()
 	case "flush-metrics":
 		metrics.FlushAll()
 	case "rollup":
@@ -240,5 +241,5 @@ func purgeDeletedFiles(days int) {
 	metrics.Event("info", "purged deleted file rows", map[string]any{
 		"kind": "maintenance", "deleted": total, "olderThanDays": days, "tookSec": int64(time.Since(started).Seconds()),
 	})
-	requestHeavyRefresh(false)
+	requestHeavyRefresh()
 }