Automatically load available models for each provider.

[Art9681]

Currently, Manifold has hard coded models for the public providers, and the local provider models are provisioned manually by the user. The exception to this is Manifold will manage a local llama-server instance with a predefined GGUF model if the config is set to single node instance. We need to improve this process. The goals are:

• Each public provider should automatically load the models available to the API key and list those in the relevant elements.
• Implement a feature to manage local models. For now, llama.cpp and ideally mlx_lm.server. Manifold should enable a way to download models for the backends in the configured datapath, and list those models when the local provider is selected in the relevant elements.