Consecutive planning_request_adapters can counteract each other

[AndyZe]

Description

We almost always begin motion planning with these two planning request adapters:

1. default_planner_request_adapters/FixStartStatePathConstraints
2. default_planner_request_adapters/FixStartStateCollision

They both "jiggle the start state" of the robot to get it back to a good state. This is just an observation that (2) can undo the effects of (1) so this strategy might not make sense.