Build llama.cpp CUDA

cmake -B build -DGGML_CUDA=ON && cmake --build build --config Release