Make llama.cpp brew install-able on Mac!

[Vaibhavs10]

Hey hey! - I'm a fellow llama.cpp enthusiast and am interested in making llama.cpp more accessible to people by making it brew install friendly!

Here's my first attempt at it: https://github.com/Vaibhavs10/homebrew-tap/blob/main/Formula/llama.cpp.rb

You can install it via brew install vaibhavs10/tap/llama.cpp

After a successful installation, you can run the llama --version from anywhere on your Mac.

But, of course, as you can see, this is far from being ready for widespread usage yet.

Couple of issues that I can already see:

1. SHA mismatches (since it clones the repo, it results in SHA mismatches whenever a change is made)
2. I get

-[MTLComputePipelineDescriptorInternal setComputeFunction:withType:]:722: failed assertion `computeFunction must not be nil.'

when passing a model file via llama -m <path to GGUF> - I am at my wit's end to figure out why that is happening.
3. It still doesn't wrap the llama.cpp server, which would be an excellent value addition, in my opinion.

I'd love to work with you all/ community here to make this possible and also make it work with the server example.

[Vaibhavs10]

Also currently facing this error when I try to load the model: #5376

I'm assuming I'm messing up somewhere while building llama.cpp in the .rb file.