Simplify transform_to_position and fix rotation denormalization

[Jondolf]

Objective

Fixes #573 (hopefully)
Fixes #618

The transform_to_position system is currently doing some rather complicated computations that appear to be prone to causing error and rotation denormalization.

It is somewhat difficult to understand what exactly the system is even doing (it has been a long time since I wrote it), but it essentially seems to apply the change between PreviousGlobalTransform and GlobalTransform, with several seemingly unnecessary extra steps, instead of simply setting Position and Rotation to match GlobalTransform.

Solution

Significantly simplify transform_to_position, and simply set Position and Rotation to match GlobalTransform.

In my testing, this fixes the denormalization issue in #618, and I suspect that it may have also helped with #573. Confirmation from @coreh or someone else struggling with the runtime panic would be appreciated!

Additional Information

I'm not entirely sure why I originally wrote this the way I did. This commit implies that it was done to fix some problem with change detection (maybe for sleeping?) but the version in this PR seems to work perfectly fine based on my brief testing.

I haven't noticed this breaking anything else, and I don't see why it would have, but if someone does notice regressions, please let me know!

[matjlars]

I tested this branch with my game and it works! Thanks again!!

[68317fa2]

I tested it too and can confirm this with a different example, where the panic, especially the whole denormalization stops after the different branch. However, I used the dynamic character example with extended mouse movement and could imagine, that this could also be somehow related to bevyengine/bevy#16480 (comment). I also want to thank you very much @Jondolf for your time and for generally maintaining this project!

[coreh]

Hey, sorry for taking this long to respond! Just revered my workaround, and updated to the latest version, and I am getting no more crashes, or the weird rotation behavior, so I believe this is indeed fixed.