Have evaluator model config work with project endpoint

[luigiw]

Is your feature request related to a problem? Please describe.
A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]

Customers usually run into the confusion of using project endpoint in evaluator model config. We should consider making the model config accept a project endpont too assuming the model deployment is in the same resource.

To be clear, the azure_endpoint in AzureOpenAIModelConfiguration(
   # can only use the azure openai endpoint. If using foundry endpoint, there will be an error
   azure_endpoint=os.environ["AZURE_OPENAI_ENDPOINT"],
   azure_deployment=os.environ["AZURE_OPENAI_DEPLOYMENT"],
)
 
must be Azure OpenAI endpoint, not Azure AI Foundry project endpoint (which is reserved for the parameter called azure_ai_project.  

Describe the solution you'd like
A clear and concise description of what you want to happen.

Describe alternatives you've considered
A clear and concise description of any alternative solutions or features you've considered.

Additional context
Add any other context or screenshots about the feature request here.