reshape op

transpose op

operator init texture itself

Metal

add openmp macro in depthwise_conv_3x3

add mobilenet mulop && add unit test for mobilenet combine  closed #745

Change op_param to support FPGA quantification close #744

change include file name close #742

update unit test and performance data of squeezenet and yolo

implement multithreading 3x3 s2 depth_conv

fix:738

Add #endif in fusion_elementwise_add_relu_op.h  close #734