Investigate using make instead of cmake to build shared library

[abetlen]

It's been pointed out that make may be better supported by llama.cpp (on some platforms). We're currently using scikit-build to build the shared library on installation with cmake but it also supports make.

Additional note as pointed out in #32 we should support passing environment variables in both settings.

[pjlegato]

With current master branches from both projects, I can get it to work by going into vendor/llama.cpp, then doing make libllama.so, then manually copying that .so to the main project's llama_cpp directory (as suggested at ggerganov/llama.cpp#797). The automatic build process doesn't work.

[abetlen]

https://cmake.org/pipermail/cmake/2010-November/040631.html suggests adding a custom target to the root CMakeLists of this project and calling make in llama.cpp. I'll try this approach.

[abetlen]

@pjlegato I've pushed a fix to build the project using make on linux and macos, do you mind testing it for me? If you can confirm that it works I'll push to PyPI.

EDIT: Got confirmation this worked, going to close this issue.