Update llama.cpp

[arrizalamin]

I've updated llama.cpp to the latest version (081fe43) and fixed the build scripts to build the llama-cli target (ggerganov/llama.cpp#7809). This added support for several new small models such as phi-3 and support for T5 and Flan-T5 model families. I've also updated the example HTML files (example-multi-thread.html and example-single-thread.html) to include these new model options in the dropdown.