Fix corruption when a URDF file is included from a SDFormat 1.6 model

[azeey]

This fixes #2728 by sending the non-parsed text of the model to gzserver. Before this, when spawning a model using the GUI or the gz tool, the SDFormat/URDF file was parsed by libsdformat on the client side for error checking or previewing model placement in the world. This was also used to set a new name for the model. The parsed model was then reserialized to xml and sent to the server in a Factory message. With this PR, the parsing still happens for the same reasons, but instead of reserializing the model, the original text in the file is sent over a Factory message. To allow setting a new name, a new field ,new_name, is added to the Factory message type.

See #2728, specifically this comment for why reserialization is a problem.

[traversaro]

Thanks a lot @azeey for working on this! I tried this branch on my machine, but it seems that the problem #2728 is still present, I wounder if I have some local problem or there are still some problems after the fix in this PR.

[traversaro]

For reference, if instead I change https://github.com/osrf/gazebo/pull/2734/files#diff-b8479763685d74e12aa126230aae05ccR112 from:

sdf::readFileWithoutConversion(_filename, this->dataPtr->modelSDFUnconverted,
          _filename, this->dataPtr->modelSDF, errors))
                                 errors);

to

sdf::readFile(_filename, this->dataPtr->modelSDFUnconverted,
          _filename, this->dataPtr->modelSDF, errors))
                                 errors);

the behaviour is:

so it seems that the preview is malformed, but the final spawned model is correct.

[azeey]

@traversaro Apparently I had set the sdf version to 1.7 in my model_sdf_1_6 model for testing and forgot to change it back. As you observed, the problem is still there.

[EricCousineau-TRI]

Per VC, Addisu has some tests that should show the fix, and then this will turn into a non-draft PR.

[azeey]

I've added tests that exercise spawning a nested URDF from the GUI as well as using the gz tool. @traversaro, do you mind reviewing the changes I made? BTW, thanks for providing the sample files.

[traversaro]

Thanks a lot, I am checking them now.

[traversaro]

The PR seems great, especially the GUI tests. However, it would be useful if the PR contained a small description about the fix (just a sentence, nothing to elaborate) as I far as I understand:

This fixes #2728 by keeping an extra model that is read in without up-conversion. This is then used for serializing and sending the model to gzserver.

Is not relevant anymore.

[azeey]

Thanks for the review @traversaro . I've updated the PR description.

[j-rivero]

Quick look at the code and added some questions.