Jan 0.6.8 update: enhanced model stability, faster CPUs, and new configuration options

Jan 0.6.8 delivers improved model loading stability and clearer error handling for llama.cpp, making troubleshooting more straightforward. The update introduces faster Mixture of Experts models on CPUs and support for searching private Hugging Face models. New support for custom Jinja templates provides advanced configuration options.