Created by: kexinzhao

fix #8816 (closed) fix #9021 (closed)

numpy float16 is internally represented as numpy.uint16. Hence, we create a binding via the help of the uint16_t type.