[FEAT] Support BrainForge LLM Backend

[NeonDaniel]

Objective

Add support for the BrainForge service backend, similar to vLLM but with specific advertised models and personas.

Initial Implementation Requirements

• Implement a backend, similar to vLLM that accesses https://hana.neonaibeta.com/docs#/llm endpoints
• Populate BrainForge expert personas from the /get_personas endpoint and display as advisors
• Support configuring a specific model to be used for Orchestrator tasks

Other Considerations

• Similar to vLLM, the models (and advisor personas) may change at any time, so they need to be sync'd and errors should be handled nicely
• Avatars might follow model name, but likely a default will be necessary since the available models may change at any time