Query-Based Adaptive Aggregation for Multi-Dataset Joint Training Toward Universal Visual Place Recognition