Add nuance to the glTF coordinate migration guide

[janhohenheim]

How can Bevy's documentation be improved?

Whew, long story. But in short, the migration guide needs to reflect the insights from https://discord.com/channels/691052431525675048/743663924229963868/1394079989552644186 onwards

• glTF uses two different systems for models and cameras/lights
• Bevy uses one for cameras and implicitly the same for models
• forward semantics were different for cameras and glTF models
• the migration rotates the imported models so that semantics are the same everywhere, at the cost of making imported cameras technically not glTF compliant as they face +Z when using an identity transform
• this should be fine, as scenes still look exactly the same in practice when exporting cameras in glTFs

[janhohenheim]

Note: ping Griffin for a review