[BUG] Azure OpenAI API chat/completions/response/object

### API Spec link

github.com/Azure/azure-rest-api-specs/blob/main/specification/cognitiveservices/data-plane/AzureOpenAI/inference/preview/2025-02-01-preview/inference.yaml#L4286-L4293

### API Spec version

2025-02-01-preview

### Describe the bug

https://github.com/Azure/azure-rest-api-specs/blob/main/specification/cognitiveservices/data-plane/AzureOpenAI/inference/preview/2025-02-01-preview/inference.yaml#L4286-L4293

```yaml
object:
type: string
description: The object type, which is always `chat.completion.chunk`.
enum: [chat.completion.chunk]
required:
- prompt_tokens
- completion_tokens
- total_tokens
```

Unexpected `required` field content for the `object` **string** field.

It seems like the schemas for `object` and `usage` fields were mixed.

### Expected behavior

`required` field is redundant and could be removed

### Actual behavior

`required` is present and confusing since the `object` field is of string type

### Reproduction Steps

Observe https://github.com/Azure/azure-rest-api-specs/blob/main/specification/cognitiveservices/data-plane/AzureOpenAI/inference/preview/2025-02-01-preview/inference.yaml#L4286-L4293

### Environment

_No response_

Open specification/cognitiveservices/data-plane/AzureOpenAI/inference/preview/2025-02-01-preview/inference.yaml and inspect lines 4286-4293, focusing on the chat completion chunk object's schema. Remove the redundant required field from the string object property, then verify that the schema matches the API specification and no longer mixes the object and usage fields.

Written by the indexing model from the issue text.