Add support for llama-cpp router mode and llama-swap

Hi, there again.

Continuation of #1 (new issue).

llama-server implemented a router mode a while ago, when the main process acts as a router starting each model as requested, listing all models configured in a separated [models.ini](https://github.com/ggml-org/llama.cpp/tree/master/tools/server#model-presets) file. This makes switching and maintaining different configurations for each model much simpler, more or less like what llama-swap does.

It would be nice if llama-panel could support both. llama-swap has a web panel, but not a separated desktop app, also as far as I know it does not support llama-server router mode by its own. So it would be nice to have a desktop app that would manage everything in a single sleek app.

Another reason for this is that llama-server also uses a separated models folder where individual GGUFs can be store, apart from HF hub. The hierarchy is explained [here](https://github.com/ggml-org/llama.cpp/tree/master/tools/server#model-sources).

These two environment variables could be used too:
LLAMA_ARG_MODELS_DIR - GGUFs dir
LLAMA_CACHE - usually point to HF hub cache (IIRIC it is how it works by default as of now)

So it would be nice is llama-panel could detect is llama-server is running in router mode when it starts and use it. This can also gracefully start and stop any model individually.