Harnessing the Power of LLMs: Automating Unit Test Generation for High-Performance Computing