Towards Fair RAG: On the Impact of Fair Ranking in Retrieval-Augmented Generation