Add upconv_7 model support

[YukihoAA]

See: #40

For more information.

Summary

waifu2x-converter-cpp is based on OpenCL, CUDA, FMA, etc. without cnn library.

which means if you want to implement upconv_7, you have to implement nn.SpaticalFullConvolution with your own code.

there is tones of cnn libraries can implement waifu2x like caffe and ncnn-vulkan.

vulkan accelerates better than hard-corded opencl so there is no reason to implement this function.

but still, waifu2x-converter-cpp is good choice to use vgg model with AMD Graphics or CPU

so if you want to implement this function, please check previous articles and try!