allow blit path for read pixels if pack->alignment isn't 1.
pack->alignment doesn't matter usually if the image is sufficiently aligned
anyway. Speeds up some tests/readrate cases by a factor of 100 (since
the pack->alignment default which noone ever bothers to change is 4).
fix bogus fb/drawable information (fixes xdemos/wincopy)
the framebuffer objects attached to drawables can have invalidate state
associated with them, since for the window framebuffer this is per-context
state and not per-fbo state. Since drivers may rely on that information
(otherwise would need to check if currently the window-framebuffer is
bound in a lot of places) fix it up in _mesa_make_current (ugly).
move drawable validation back to where it was now that the driDrawable
information isn't used directly anymore.
Fix bogus fb updates (the context we get for SwapBuffer processing
may not have the drawable being processed attached!)
glthreads behaves a bit better but still not correct.
don't use driDrawable information directly, don't resize at makecurrent
driDrawable information now isn't used in most parts of the driver.
makecurrent only updates fb information when the context is new
(driDrawable is still updated by the dri common code).
still doesn't work quite right (resize).
Fake frontbuffer doesn't copy in real frontbuffer.
Don't even think about doing rotation/page flip/triple buffering for now...
More cleanups needed (fake cliprects etc.)