Our current reset code is a mess, we can absolutely annihilate gyroFix and leave it to yawFix to handle the yaw offset. The yaw calculation for isolating attachment pitch/roll is now using YXZ euler angle order so that roll does not affect the yaw (at the sacrifice of pitch, but it's one I'm willing to make). This also fixes changes to manual mounting orientation causing your calibration to be annihilated.

I tested this lightly and it works for IMUs, unsure about with HMD or computed trackers. I left most other jank the same because I don't want to break things, but there is a lot that can be improved. Notably, the way we handle computed trackers (which btw only means that it doesn't need mounting) is kinda weird and tposeDownFix is super weird and jank?

I would also like to make more unit tests, I spent way too much time figuring out all the logic behind the reset code so I'm prepared to actually write some new ones for computed trackers and maybe things like t-pose.

Replaces #1321