M$^3$oralBench: A MultiModal Moral Benchmark for LVLMs