Add imu 2d rostest

[efernandez]

• Add diffbot, similar to the one in https://github.com/ros-controls/ros_controllers/blob/noetic-devel/diff_drive_controller/test/diffbot.xacro, but a bit simpler and with an IMU
• Add IMU test that shows the relative constraints created don't match the ground truth when differential: true

$ make run_tests_fuse_models_rostest_test_imu_2d.test
[ROSTEST]-----------------------------------------------------------------------

[fuse_models.rosunit-imu_2d_test/BaseFrameYawAngularVelocity][FAILURE]----------
/home/efernandez/dev/ws/cpr_ws/src/fuse/fuse_models/test/test_imu_2d.cpp:255
The difference between w and orientation_delta / dt is 0.96695215295635117, which exceeds 1.0e-2, where
w evaluates to 0.48358168580225275,
orientation_delta / dt evaluates to -0.48337046715409837, and
1.0e-2 evaluates to 0.01.
--------------------------------------------------------------------------------


[fuse_models.rosunit-imu_2d_test/BaseFrameYawAngularVelocity][FAILURE]----------
/home/efernandez/dev/ws/cpr_ws/src/fuse/fuse_models/test/test_imu_2d.cpp:306
The difference between w and delta[2] / dt is 0.96695215295635073, which exceeds 1.0e-2, where
w evaluates to 0.48358168580225275,
delta[2] / dt evaluates to -0.48337046715409793, and
1.0e-2 evaluates to 0.01.
--------------------------------------------------------------------------------


SUMMARY
 * RESULT: FAIL
 * TESTS: 1
 * ERRORS: 0
 * FAILURES: 1

Some tolerances in the test are a bit too large. I believe that's due to the fact I didn't synchronize the transactions generated by the Imu2D sensor model and the ground truth. However, I don't think that's really important, because the error is quite large with the current implementation.

I've fixed this in #268

This is on top of #266, which was required for the rostest to compile because a header was missing in the Imu2D sensor model header file.