Created by: guru4elephant

currently, client does not support multi thread prediction, but users can do multi process prediction based on python subprocess. we need to wrapper this as a simple function. https://github.com/PaddlePaddle/Serving/pull/156