LLM-optimizer: Benchmark and optimize LLM inference across frameworks with ease