ColladaExporter produces 6 nodes for each node in the original .dae

[Blast545]

Environment

• OS Version: Ubuntu 20.04
• Source build

Description

• Expected behavior:
  Generated output dae file with 1 node for each node in the input dae file.
• Actual behavior:
  Generated output with repeated nodes.

Steps to reproduce

Working on top of #763 (mainly to have available the test file and submeshes fix).

On top of the environment files, I'm running:
ign gazebo -v 3 -s -r --iterations 1 src/ign-gazebo/test/worlds/collada_world_exporter_submesh.sdf

to get a exported .dae with the contents of the original tunnel_tile_6. I would expect the final result to have as many nodes as the original .dae file. However, the final result in this scenario has 6 nodes for each original node. They repeat in regular steps, that is, in the final result file:

node_0, node_5, node_10, node_15, node_20, node_25 are equivalent.
node_1, node_6, node_11, node_16, node_21, node_26 are equivalent.
And so forth.

Adding a video to show the problem:

bug_exported_nodes.mp4

[Blast545]

The main cause of this issue is related to the fact that models in fuel like this one: Tunnel Tile 5 have in their model.sdf the meshes loaded once for a main visual and later the mesh is loaded several times, once per each sub-mesh. Attaching the model.sdf file of that model for reference (in.txt format, github reasons).
model.txt

#802 Addresses the submeshes being exported as the complete mesh of the file, now the only part exported is the submesh marked with the sdf tag .

I think the ObjLoader or the ColladaLoader shouldn't be importing the whole mesh file if a submesh is selected, although this can be considered a separated issue.

[Blast545]

I just realized the MeshManager is able to load meshes only once per file here, so I don't think there's an important issue loading the whole mesh.