Use dynamics model in state estimation

[kev-the-dev]

Summary with wise words from MIL alum during a discussion about helping the Sub's state estimation problems when DVL beams do not return:

Most teams that do not have a DVL achieve useful state estimation by using a model of their vehicle's dynamics as a predictor in their kalman filter. An example of this is modeling the vehicle's drag/buoyancy/gravity and feeding the acceleration you would expect from these forces + the current wrench into the kalman filter.

May be related to #380, which will model the Sub's dynamics too for simulation purposes.