Add support for Github models (models.inference.ai.azure.com) API

[aenold]

Is your feature request related to a problem? Please describe:

Currently, bolt.diy does not directly support the Github Models API available at https://github.com/marketplace/models/ with the endpoint https://models.inference.ai.azure.com. This limits the ability to integrate models hosted on this platform, specifically models like the o1-preview and other Github models. This would be beneficial in extending the scope of models the application has access to.

Describe the solution you'd like:

I would like bolt.diy to include a new provider configuration and associated logic to interact with the Azure AI Inference API at https://models.inference.ai.azure.com. This would involve:

• Adding a new provider called "Github models" or similar.
• Supporting authentication with API keys.
• Adding the available models to the MODEL_LIST.
• Adding the https://models.inference.ai.azure.com endpoint to the list of valid providers.
• Support passing an model name to the api using client.chat.completions.create

Describe alternatives you've considered:

As a workaround, we could attempt to use the "OpenAILike" provider. However, this is not ideal, it might not fully support all features specific to Azure AI Inference API, and would not provide an ideal experience for users.

Additional context:

The Github models API at https://models.inference.ai.azure.com offers access to models like o1-preview and Llama-3.3-70B-Instruct and could broaden the models available through bolt.diy. Having direct support would improve the user experience for anyone looking to use models from Azure. This would allow bolt.diy to support a bigger array of models.