Remove DescartesVertexEvaluator which duplicates descartes::StateEvaluator

[marip8]

The tesseract::DescartesVertexEvaluator is a class that is confusingly similar to the descartes::StateEvaluator and is used to do effectively the same task as the descartes::StateEvaluator but inside the tesseract::DescartesRobotSampler class. This is not really necessary because states will get pruned within the descartes solver itself before edges are added (here fi).

To eliminate this confusion and additional unnecessary class, my suggestion is that we:

• port tesseract::DescartesJointLimitsVertexEvaluator to be a descartes::StateEvaluator
• remove the concept of the DescartesVertexEvaluator.

Also, we should not be adding arbitrary vertex evaluators to the robot sampler when the user chooses a nullptr for their vertex evaluator selection:

tesseract_planning/tesseract_motion_planners/descartes/include/tesseract_motion_planners/descartes/impl/profile/descartes_default_plan_profile.hpp

Lines 205 to 219 in fcaf317

	std::shared_ptr<descartes_light::WaypointSampler<FloatType>> sampler;
	if (vertex_evaluator == nullptr)
	{
	auto ve = std::make_shared<DescartesJointLimitsVertexEvaluator>(prob.manip->getLimits().joint_limits);
	sampler = std::make_shared<DescartesRobotSampler<FloatType>>(mi.working_frame,
	cartesian_waypoint,
	target_pose_sampler,
	prob.manip,
	ci,
	mi.tcp_frame,
	tcp_offset,
	allow_collision,
	ve,
	use_redundant_joint_solutions);
	}

This should become a non-issue if we remove the class.

[Levi-Armstrong]

I do not have objections to your proposed changes.