I don't know enough yet about Construct to comment on this for sure (still really new to Construct), but if I had to guess, I'd think that option 1 would be slightly better. I say this because I would *guess* that Construct does not create all objects in memory from all layouts at the start. Therefore you have fewer objects in memory at a time by splitting into separate layouts. If you have both layers in one layout, all of the objects in both layers are in memory at the same time.
Also there would likely be a CPU advantage to option 1.
Like I said, I don't know the inner workings of Construct yet, and the difference is likely to be small from a memory standpoint anyway, but that is what I would think...
Also it depends on the situation of course. If it is important that there be minimal delay when "switching," layers would be the way to go.