Describe the feature or improvement you are requesting

Summary
The OpenAI .NET library currently provides protocol methods for direct access to the REST API with "binary in, binary out" functionality. However, these protocol methods only support non-streaming operations. This feature request proposes adding streaming support for protocol methods to enable compatibility with other LLM providers and custom request handling in streaming scenarios.

Proposed Solution
Add a streaming variant of the protocol method:

public virtual AsyncCollectionResult<BinaryData> CompleteChatStreamingAsync(BinaryContent content, CancellationToken cancellationToken = default)
{
    var options = cancellationToken.ToRequestOptions(streaming: true);

    return new AsyncSseUpdateCollection<BinaryData>(
      async () => await CompleteChatAsync(content, options).ConfigureAwait(false),
      (jsonElement, _) => BinaryData.FromString(jsonElement.GetRawText()),
      options?.CancellationToken ?? CancellationToken.None);
}

This method has been tested successfully with multiple major LLM providers including Gemini, Claude, DeepSeek, Qwen.

If this approach is suitable, I'm ready to submit a PR with the implementation.

Additional context

No response