[Feature] Add llama.cpp vulkan backend support for non-nvidia cards