Standardize API Output Formats

[mattjhawken]

Token streaming, simple generates, and chat completions should have an up-to-standard json output format.

See tests/test_model_api for usages and ways of testing. Most of the changes would need to be made to the api/node, api/models, and ml/validator.

Ideas

• Ensure that the output and input format args perform accordingly
• Add any relevant missing input fields to models
• Reasoning blocks can be separated into a separate field for openai responses, e.g.:

 {
   "choices": [{
     "message": {
       "content": "Ahoy there! ⚓🏴‍☠️ How can I assist you today?",
       "reasoning": "<think>...</think>..."  // Optional, separate field
     }
   }]
}