From Handcrafted Features to LLMs: A Brief Survey for Machine Translation Quality Estimation