Performance improvements, Vblank support and bugfixes

Note that DRI3WSEGL without the Vsync patch (dri3wsegl!4) will judder once this is merged, so both should be merged at around the same time.

This does unfortunately make tearing in some situations slightly worse for some reason, but it was hardly great to begin with and tearing-free X11 is not such an easy task. It brings the benefit of Vsync support in 3D stuff (otherwise some emus run at 150fps the whole time) and reduced tearing in other situations (particularly fullscreen).

Soon I should have another merge request for atomic page-flip support, that will make properly-written fullscreen 3D stuff 100% tearing-free and performance-optimal (no unnecessary blits as the SGX buffer is swapped with the scanout buffer) if you disable Marco soft-compositing.

Longer term the way to eliminate tearing in desktop situations would probably be to move to Wayland and/or implement triple buffering in this driver.