support model_armor with disabling vertexai

I would like to pass model_armor_config to the Generative Language API call, and it appears that the code below is currently the only way to invoke the API while including model_armor_config. Is there a way to make this work with the vertexai=False flag, to ensure that the request is sent to the Gemini API platform rather than Vertex AI?
```
client_options = {"api_endpoint": "us-east4-aiplatform.googleapis.com"}
client = aiplatform_v1beta1.PredictionServiceClient(client_options=client_options)

request = aiplatform_v1beta1.GenerateContentRequest(
model="gemini-2.0-flash-001",
contents=[content.Content(role="user", parts=[content.Part(text="Hello world")])],
model_armor_config={
"prompt_template_name": f"projects/{PROJECT_ID}/locations/{LOCATION}/templates/{TEMPLATE_ID}"
}
)
```

首先跟踪 Python 客户端中的 vertexai=False 路径和 Generative Language API 调用，然后将其与 PredictionServiceClient 和 GenerateContentRequest 进行比较。验证在面向 Gemini API 平台时是否可以传递 model_armor_config，并确认变更完成后请求确实到达预期的平台。

由索引模型根据 Issue 内容生成。