Fix problems with QuantizedMeshLoader-generated glTFs

[kring]

Before this PR, the glTFs generated from quantized-mesh tiles were technically invalid. Our game engine clients weren't too fussed about these problems, which is why we didn't notice.

Fixed the following problems:

• The asset.version field was not set.
• The byteLength of the buffers was not set. Most clients just used cesium.data.size() anyway.
• The target type of the bufferView used to store indices was incorrectly set to ARRAY_BUFFER instead of ELEMENT_ARRAY_BUFFER.
• The bufferView used to store indices had its byteStride property set, which is not allowed.
• The glTF did not have any scenes.

[csciguy8]

Do we have any unit tests for QuantizedMeshLoader::load? (an unsurprising question from me)

If not, sounds like it would be useful. Could mimic what a typical native client would do, and have all our assertions about what a proper glTF should look like.

[kring]

Do we have any unit tests for QuantizedMeshLoader::load?

Yes we do, though it looks like it missed a rename somewhere along the way and was named TestQuantizedMeshContent.cpp instead of TestQuantizedMeshLoader.cpp. I've now renamed it in this branch.

Could mimic what a typical native client would do, and have all our assertions about what a proper glTF should look like.

That would be a glTF validator. It would be great to have such a thing in cesium-native, but it's way outside the scope of this PR.

I've updated this PR to add some assertions about the things that were fixed, though. This is really just closing the barn door after the horse already bolted, and does nothing to prove the generated glTFs aren't incorrect in other ways we haven't already noticed. But it's easy to do and perhaps could catch regressions in the future, so why not?

[j9liu]

Looks good @kring ! I'll wait for CI to pass, then merge.