[ Feature ] Support local Gemini Nano

[rons4]

Support local Gemini Nano via SDK on supported devices like Pixal 10 Pro.

[shubham0204]

@rons4 The app's logic is tightly coupled with llama.cpp bindings and assumes that only GGUFs model will be downloaded and executed. Adding support for Gemini Nano will require integration with ML Kit GenAI APIs also adding to the raw download size of the app.

This feature will be helpful for users owning devices that support Gemini Nano, hence keeping the ticket open.

[Hatchet21]

Please don't integrate external remote APIs for SLM/LLMs requiring sending data to cloud servers...for privacy reasons. The reason I'm using SmolChat is that I can use it on my owm terms, in a privacy respecting way - without sending my data to any cloud servers and run models locally. If you might have to add support for APIs like that please make it optional and disabled by default.