GreedLlama: Performance of Financial Value-Aligned Large Language Models in Moral Reasoning