Towards Trustworthy Recommender Systems: From Shallow Models To Deep Models To Large Models