LiteLLM ignores GOOGLE_GENAI_USE_VERTEXAI=FALSE environment variable and defaults to Vertex AI

### What happened?

When GOOGLE_GENAI_USE_VERTEXAI=FALSE is set, LiteLLM should use Google AI Studio with the GOOGLE_API_KEY, not attempt Vertex AI authentication.

LiteLLM defaults to Vertex AI regardless of the GOOGLE_GENAI_USE_VERTEXAI setting, causing authentication errors:

I have fixed the issue I will be raising the PR soon

### Relevant log output

```shell
litellm.exceptions.AuthenticationError: geminiException - {
"error": {
"code": 400,
"message": "API key not valid. Please pass a valid API key.",
"status": "INVALID_ARGUMENT",
"details": [
{
"@type": "type.googleapis.com/google.rpc.ErrorInfo",
"reason": "API_KEY_INVALID",
"domain": "googleapis.com",
"metadata": {
"service": "generativelanguage.googleapis.com"
}
}
]
}
}
```

### Code of Conduct

- [x] I agree to follow this project's Code of Conduct

Look for the Google AI/Vertex AI client initialization in the LiteLLM codebase, likely in a provider-specific module. The environment variable GOOGLE_GENAI_USE_VERTEXAI should be read to decide between Google AI Studio and Vertex AI endpoints. Check the authentication flow and ensure the API key is used correctly when the flag is false. Verify the fix by setting the environment variable and testing the Gemini API call.

Escrita pelo modelo de indexação a partir do texto da issue.