Distortion initialization fix with background color

[WilliamLewww]

Closes #2527 - Last camera's distortion affects previous cameras' distortions

When setting the background color to a scene, the distortion parameters continue to leak. From #3033 we can infer that setting the scene's background color is effecting the distortion compositor which is causing the improper order for compositor initialization.

To fix the issue, we reinitialize the distortion after setting the scene's color to ensure the distortion compositor is properly set for each camera.

[scpeters]

this has merge conflicts that need to be resolved

[scpeters]

I tightened some test expectations in 0fe817c

[scpeters]

for some reason, there is a regressions in CameraSensor.CheckNewAndLegacyDistortionModes:

• https://build.osrfoundation.org/job/gazebo-ci-pr_any-ubuntu_auto-amd64-gpu-nvidia/760/testReport/junit/(root)/CameraSensor/CheckNewAndLegacyDistortionModes/
• https://build.osrfoundation.org/job/gazebo-ci-pr_any-homebrew-amd64/2520/testReport/junit/(root)/CameraSensor/CheckNewAndLegacyDistortionModes/

I added a getchar() statement to the end of the test so I could open a gzclient instance and view the images generated in the test. It appears that with this pull request, the distortion of the pincushion_legacy camera images has changed (all other images appeared unchanged).

Since the pincushion_legacy camera is the first one spawned in the test, I tried changing the order in which they are spawned, and I noticed that the distortion of the first spawned camera seems to be different than if it was spawned later. I'm still debugging this, but I think it's something we need to fix

[scpeters]

for some reason, there is a regressions in CameraSensor.CheckNewAndLegacyDistortionModes:

• https://build.osrfoundation.org/job/gazebo-ci-pr_any-ubuntu_auto-amd64-gpu-nvidia/760/testReport/junit/(root)/CameraSensor/CheckNewAndLegacyDistortionModes/
• https://build.osrfoundation.org/job/gazebo-ci-pr_any-homebrew-amd64/2520/testReport/junit/(root)/CameraSensor/CheckNewAndLegacyDistortionModes/

I added a getchar() statement to the end of the test so I could open a gzclient instance and view the images generated in the test. It appears that with this pull request, the distortion of the pincushion_legacy camera images has changed (all other images appeared unchanged).

Since the pincushion_legacy camera is the first one spawned in the test, I tried changing the order in which they are spawned, and I noticed that the distortion of the first spawned camera seems to be different than if it was spawned later. I'm still debugging this, but I think it's something we need to fix

I've added a test to detect the problems with the first distorted images in edd5d43

[WilliamLewww]

The tests are working now! Turns out that the distortion effect was magnified because refreshing the compositor just adds another compositor on top of the existing one. To fix the issue, I remove the compositor right before adding the new one.

[scpeters]

The tests are working now! Turns out that the distortion effect was magnified because refreshing the compositor just adds another compositor on top of the existing one. To fix the issue, I remove the compositor right before adding the new one.

great work!

[scpeters]

I noticed you relaxed some test tolerances in ff32b82 by necessity; I've tightened them back a bit in 4d284d7

I also squeezed in some fixes for clang warnings in c6d3bb9 and e47c87e and merged with gazebo11 to fix the ABI checker build