Release b188

feat: Support stream usage statistics

• Added stream_options to ChatCompletionRequest and CompletionRequest models.
• Automatically inject stream_options: {"include_usage": true} in build_request_body_generic for streaming requests if not provided.
• Enhanced stream_logger_task to parse timings from llama.cpp responses and convert them to standard OpenAI usage format for logging.