[ GPU ] split kernel registration from forwarding function in addition_layer_cl and transpose_cl

[ GPU ] split kernel registration from forwarding function in addition_layer_cl and transpose_cl #607