add refact llama.cpp tutorial

[ds5t5]

/claim #77

reference PR: https://github.com/ggerganov/llama.cpp/pull/3329

[ds5t5]

@olegklimov please help review and feel free to test. The inference is extremely fast with the effort from llama.cpp.

[teleprint-me]

@ds5t5

It is really fast! Nice work!