From Tokens to Faces: Investigating Discrete Speech Representations for 3D Facial Animation — ThinkLLM