Are Language Models Sensitive to Morally Irrelevant Distractors?