This layer enables user to write conv layer and fc layer in one nn.Sequential model.

My implementation is at here:
https://gist.github.com/VoVAllen/5531c78a2d3f1ff3df772038bca37a83

Quite simple, but convenient for computer vision beginners.