More GLX janitation

All small and hopefully obvious. The first two clean up some aspects of how we handle the GLXGetDrawableAttributes request. The third I encountered while trying to debug no-config contexts, we were synthesizing an error for MakeContextCurrent even though the app never called the two-drawable form and the protocol tries hard to use the one-drawable form if they're equal, which was very confusing. The remainder lifts the common logging code from the dri backend to glx and tries to use it consistently instead of fprintf.