Optimizing the general movement in "3d" game is a problem is most of the time a problem with two optimization parameters (X and Y coordinate), sometimes a third parameter is there, e.g. when there's non-constant jumping height or flying freely on all 3 axises.
Going from one place to another is an easy problem, even with an crazily rotating camera. Most of the time you only need to work on one frame, it's very unlikely that a frame will saved by going into a suboptimal angle initially or manually rotating the camera. Since most TASers only use the very outer range of the analog stick to adjust angles, their work is not as optimized as it could be. Using all coordinates on the analog stick doesn't make it any harder, you just use
a line drawing algorithm on the analog stick to calculate the next x,y pair.
While going around corners you need to test two scenarios: let speed drop to go in a tighter angle or don't let speed drop. Both can be done by the same algorithm as above. For the latter you try the most tight angle that doesn't drop speed, for the first you need to decide when an how much speed should be dropped. You should consider the level layout as well when deciding which approach you want to make.
