* LLM: add llama2-32K example. * refactor name. * fix comments. * add IPEX_LLM_LOW_MEM notes and update sample output.