Efficient Continual Learning in Neural Machine Translation: A Low-Rank Adaptation Approach