Integrate llama.cpp

[jorgeantonio21]

It will likely that we will want to integrate llama.cpp (or one of its available rust bindings) to our stack. It will be important to have comparison benchmarks. The following is required

• Benchmark llama.cpp vs hugging face candle. Also, map which model architectures are available for each of these libraries.
• (Optional) Create our own rust bindings library around llama.cpp.
• Integrate llama.cpp directly into our tech stack.
• The user should be able to compile the project either with llama.cpp or candle, not both. This can be achieved through a suitable feature configuration.

[jorgeantonio21]

Not relevant atm.