Use makefiles for build

[aittalam]

Moving the llama.cpp to makefiles to simpify integration with the rest of llamafile's code.

• added BUILD.mk so we can do without cmake
• build works with cosmocc 4.0.2
• dependencies are all taken from llama.cpp/vendor directory
• building now works out of the box both on linux and mac

see README_0.10.0.md for further updates

[peteski22]

Amazing @aittalam 👍🏼

I added a couple of comments, and I'd just add we (I can help) should test this on Linux as well as MacOS before we merge anything to main later on.

Thanks! 😄

[aittalam]

Amazing @aittalam 👍🏼

I added a couple of comments, and I'd just add we (I can help) should test this on Linux as well as MacOS before we merge anything to main later on.

Thanks! 😄

Thank you! 🙏 I agree, I have tested this on linux too but we should make sure this is tested automatically every time: we do have code to do that on CI, I just have broken the build process atm 😅 but we'll make sure it is ok before we merge the new_build_wip branch with main