Check restrictions in addto_layer

[guoshengCS]

Must addto_layer check the height, width, depth attributions of the inputs? Might it be better to remove the check of height, width, depth in addto_layer for common numerical calculations besides image processing? Besides, even image processing network configurations might fail since some layers remove the height, width, depth attributions.
Additionally, it seems that the check makes addto_layer only support two inputs, while it actually supports adding more than two inputs.