Add ability to retry to other model if 429 is received

[adilhafeez]

If a model returned 429 then try other configured model. For example if 429 is returned when accessing gpt-4o then llm_gateway should retry to other model for example in this case we should retry with gpt-5.

...

model_providers:
  - model: openai/gpt-4o
    base_url: https://dsna-oai.openai.azure.com
    access_key: $OPENAI_API_KEY
    retry_on_ratelimit: true # new feature

  - model: openai/gpt-5
    base_url: https://dsna-oai.openai.azure.com
    access_key: $OPENAI_API_KEY

...

We could add add another field retry_on_ratelimit which if set will allow plano to retry to other models.

[salmanap]

I think we should establish retry mechanism (exponential, linear, etc) and then have a fallback = model also get configured