Fine-Tuning LLaMA-3 for Psychology Question Answering Using LoRA and Unsloth