Enable computation of sparse DRRs

[eigenvivek]

Want to be able to compute just a subset of the pixels

also would be helpful to have sparse implementations of loss functions too!