Streamlined line sensor model.

We now use a single position for all pixels (but of course still
different line-of-sight vectors), as it doesn't really make sense to
consider different positions as pixels are only a few microns away from
each other.