IMU Yaw Mesurement in Odometry

[jgueldenstein]

The yaw estimated by the IMU could be used in the odometry that feeds into the localization. This could help us to prevent cases where the robot is turned around (like the game against NuBots) on the center circle.
IMU yaw drift should be less than the error from the walking odometry.

Care must be taken as the IMU orientation measurement is already used in the localization when the robot falls. This may need to be disabled to not use it twice.

[val-ba]

Handled in #623

[Flova]

The PR was reverted in #667 because the old IMU is not good enough to support this. This would need to wait for the new one. Then the revert can be reverted