Visualization collada texture in marker fix

[mcopejans]

Fixes COLLADA mesh material colors as described in issue #751. From the issue thread, it looks like this case has already been fixed by @wjwwood (752), but unfortunately it got unfixed by another PR (893).

• Tested locally with Collada file generated by Blender. Without the fix, this image would be plain white.
  
• Tested with example by @garaemon: https://github.com/garaemon/rviz_collada_marker
• The fix implemented here was originally proposed by @brindza

[wjwwood]

Thanks for the pr, especially since I haven't had the time to look into the reports of this:

e5ecd45#comments

I feel a bit like a leaf in the wind on this issue, since I don't have a good way of sorting out which solutions to merge or not, which probably led to the "unfix" of this issue. What I'm missing from this is confirmation that we're not "unfixing" something else with this pull request. I'd really like some feedback from @Logane or @brindza or even @hersh.

Otherwise, I'll have to sit down and try to run the "demos" we have to see that we're not breaking some other expected behavior. That will probably take some time for me to get to myself. If anyone else has time to look at the demos with this pr and corroborate with @mcopejans that we're making purely forward progress I'd appreciate and I could merge this directly. Specifically I'm talking about the demos in these images:

#893 (comment)

@mcopejans I don't mean to discount the testing you've done either. Perhaps some before and after images of the tests you did would help too:

Tested with example by @garaemon: https://github.com/garaemon/rviz_collada_marker

[mcopejans]

@wjwwood Thanks for your time, here are some before-after images. Let me know if there is anything I can do to make sure that this PR does not break something else.
Before

After

[wjwwood]

I've been doing some testing with this pull request, and it's a definite improvement based on the https://github.com/garaemon/rviz_collada_marker demo. I couldn't find any cases where there is an obvious regression. So I'm going to merge this in the hope it's two steps forward and none backwards.

Thanks @mcopejans for the pr and sorry it took me so long to get to it.

I'm going to forward port this to jade, kinetic and lunar too.