Include ExternallyAppliedSpatialForce input in KinematicTrajectoryOptimization::AddEffortBoundsAtNormalizedTimes

[aelvio]

The implementation of KinematicTrajectoryOptimization::AddEffortBoundsAtNormalizedTimes, internally, assesses generalized forces by:

MultibodyForces<AutoDiffXd> forces_(*plant_); plant_->CalcForceElementsContribution(*plant_context_, &forces_); *y = plant_->CalcInverseDynamics(*plant_context_, qddot, forces_);

However, this does not account for any additional forces that may be acting on the system outside of a ForceElement, such as contact forces or externally applied spatial forces (which I'm using to mimic contact forces).

Ideally, fixing the value of the plant's applied spatial force input in a particular context should get that spatial force considered in the EffortBounds calculus (and perhaps elsewhere).

@RussTedrake mentioned that it would be good to support this (and perhaps make it easier to support in all of the control methods which evaluate the dynamics). I don't know if contact forces would introduce other issues, but at the very least it would be great to have externally applied spatial forces included.