feat(ai-core): trace instrumentation, LLM thinker, decay scheduler

- orchestrator: Tracer interface with spans for intent analysis, persona,
  subsession dispatch, context build, LLM streaming and each tool call;
  traces finish when the stream completes — the admin Trace waterfall
  finally receives data
- background: Thinker generates real LLM thoughts (post-chat + silence
  triggers, recent history as context, empty/overlong outputs discarded);
  thoughts broadcast to web clients as 'thinking' WS messages; pipeline
  postChatHook triggers post-chat thinking after each round
- memory: ApplyDecayAll + ListUserIDs with a 6h decay/purge cron in main
  (was implemented but never scheduled)

backend/services/memory/repository.go

@@ -221,6 +221,24 @@ func (r *Repository) RecordAccess(ctx context.Context, id string) error {
 	return err
 }
 
+// ListUserIDs returns distinct user IDs that have memories.
+func (r *Repository) ListUserIDs(ctx context.Context) ([]string, error) {
+	rows, err := r.db.QueryContext(ctx, `SELECT DISTINCT user_id FROM memories`)
+	if err != nil {
+		return nil, fmt.Errorf("list user ids: %w", err)
+	}
+	defer rows.Close()
+	var ids []string
+	for rows.Next() {
+		var id string
+		if err := rows.Scan(&id); err != nil {
+			return nil, err
+		}
+		ids = append(ids, id)
+	}
+	return ids, rows.Err()
+}
+
 // PurgeDecayed removes memories below the decay threshold.
 func (r *Repository) PurgeDecayed(ctx context.Context, userID string) (int64, error) {
 	result, err := r.db.ExecContext(ctx,