Llama.cpp Vulkan support

[atirut-w]

Using Llama.cpp with Vulkan support should provide GPU acceleration on a wide range of graphics cards. It is also easier to integrate than Ollama. For a good example, see LM Studio.

[FrancescoCaracciolo]

LM studio supports OpenAI like API, so you can use the OpenAI Handler and running it externally, if it solves your use case.
Maybe we can add a specialized handler that can do a few other things like for ollama.

For direct llama.cpp support, getting hardware acceleration work under Flatpak always requires more effort. I will check what I can do