add visual odometry residuals/formulation

[navganti]

Currently we have no development for visual odometry. This may include some modifications to the vision front end, and will definitely require development of the vision back end.