New OpenGL threading code - call for testing

[Henrik]

I've implemented (with a bit of help from Unknown) threaded rendering for OpenGL, which means that the main thread queues up commands, and a background thread actually sends them to the driver, similarly to what we've been doing with Vulkan for a while. This arrangement also allowed more efficient vertex uploads, which improves performances further, but Vertex Cache should no longer help very much.

With this, you should be seeing much better performance if your device couldn't quite keep up before. There is no option, it's always on in new builds.

Please try the latest builds from https://buildbot.orphis.net/ppsspp/ and let me know if anything is broken graphically that used to work, either here in this thread, or preferably in the issue tracker: https://github.com/hrydgard/ppsspp/issues