FairQueue: Rethinking Prompt Learning for Fair Text-to-Image Generation