Support reasoning generation property

[gerukin]

Is your feature request related to a problem? Please describe.
Reasoning models output both a final answer and a reasoning (chain of thought) part. The format depends on the provider and model. This Ollama provider does not seem to currently allow this for any model, requiring manual parsing to separate the reasoning from the actual answer.

Describe the solution you'd like
Some models and providers (ex: R1) support a reasoning property on the generated response object. It would be ideal if this were automatically set based on the selected model.

Describe alternatives you've considered
It is possible to use extractReasoningMiddleware from the AI SDK, but this requires keeping track of model specifics (like model name regex and tag) in the app. This is more of a fallback solution, and normally not required for well supported models in other providers.

Additional context

• N/A -