I am by far not an expert at this, for I sometimes wonder myself. I do know that a normal map does not displace geometry or even appear to displace it. That, I believe, is what a voxel map does. A normal map is like a bump map in the fact that it makes the illusion of depth. It is different from a bump map because it also gives information to the light and shadow rays to make it look like shadows and highlights are falling into that depth. Perfect example is place a bump map on a plane and look at it at a low angle. It will look smooth even though you can see that illusion of depth. Now place the normal map on that same plane and it won’t look smooth anymore, but like you have geometry for all that depth in there.
Now, from what I’ve read and understood (which is probably wrong), is that the displacement socket can only give you bump map results and the normal sockets can give you either depending on the image you plug into it.
For your other question, a bump map is usually a black and white image with white being highest points (value of 1) and black being the lowest (value of 0). That being said, a bump (displacement socket) can use any image to create that illusion. All it does is take the color values and use that for depth.
Hope this helped a little. Sorry for the kindergarten explanation, but I’m not as learned on the subject as I should be. One who knows this very well is Sebastian Konig if you can get ahold of him.