The high fidelity of scene representation in visual long-term memory