Change default wrap mode for osg::Texture from GL_CLAMP

[AnyOldName3]

The default texture wrapping mode for osg::Texture is GL_CLAMP. GL_CLAMP was removed in OpenGL 3 as nearly every time it was used, it was used because someone thought they were getting GL_CLAMP_TO_EDGE, which is more useful. Also, GL_CLAMP tended not to be implemented with fixed-function hardware once programmable GPUs came along, and was instead shimmed in with a driver-generated shader variant. This meant it was slower, too. For context so everyone's on the same page, GL_CLAMP includes a half-texel-wide region around the texture where the border colour is blended in, whereas GL_CLAMP_TO_EDGE uses the edge texel colour in this region.

A recent discussion between the OpenMW and Zink teams revealed that OpenMW had unintentionally been using GL_CLAMP quite a lot, leading to stutters as new pipelines needed creating immediately before use. We narrowed this down to textures with no explicit wrap mode set defaulting to GL_CLAMP, and have done some work to ensure we always explicitly set something. As we use a fork of OSG, we're probably going to end up changing the default there to an invalid value, so if we forget in the future we'll get OpenGL errors. The alternative would be to set another valid default, but that would mean we wouldn't get the benefit when packagers elected not to use the fork.

I suspect OpenMW won't be the only project to have fallen for this, though. That would mean it might be beneficial for upstream OSG to make a change, too. As the upstream project, it doesn't need to mess around with the invalid default idea, and can pick a better alternative. GL_CLAMP_TO_EDGE would be one option, as that's almost certainly going to give the behaviour people who looked and saw it was GL_CLAMP were expecting, and GL_REPEAT would be another, as that would match the OpenGL default. This would technically be a breaking change, but I'd expect anyone in the rare situation of actually wanting GL_CLAMP would be a tiny minority, and likely to be outweighed by a much greater number of users whose applications such a change would improve.