URDF parsing: contents of visuals in gazebo extension gets lumped to other visuals

**Original report ([archived issue](https://osrf-migration.github.io/sdformat-gh-pages/#!/osrf/sdformat/issues/58)) by John Hsu (Bitbucket: [hsu](https://bitbucket.org/%7B0a186eae-abf0-4514-a951-23db5eccc286%7D/), GitHub: [hsu](https://github.com/hsu)).**

----------------------------------------
For example, if I had an urdf with the following extension, to try and sneak sdf code into gazebo:
~~~















0.05 0 -0.05 0 0 0


0.4 0.175 0.125




~~~
The sdf's [urdf parser lumps contents of visuals in extension](https://github.com/osrf/sdformat/blob/0029bebd3ed310d3f2f56eaf1a3166f9f582f430/src/parser_urdf.cc#L1164) into the existing visual element of the urdf, resulting in a visual block that is a combination of the urdf visual and the sdf visual in the extension:
~~~

0 0 0 0 -0 0


1 1 1
model://atlas_description/meshes/l_foot.dae


0.05 0 -0.05 0 -0 0


0.4 0.175 0.125



~~~
Commenting out [this bit of special case code for `visual`](https://github.com/osrf/sdformat/blob/0029bebd3ed310d3f2f56eaf1a3166f9f582f430/src/parser_urdf.cc#L1164) fixes this issue, but probably breaks something else.

We need a better way of adding sdf to models specified in urdf. Especially when new things are introduced in sdf, pass it through from urdf in a more generic fashion.