Use Cogl read-pixels for secondary output CPU copy
The main point of this series is to fix performance on DisplayLink outputs on a particular laptop where it takes 5 seconds to copy one full-HD frame. With patch
renderer/native: use cogl for CPU copy path that goes down to below 20 milliseconds. Support for DisplayLink outputs in the first place requires !237 (merged), but this MR does not depend on it.
The other patches are preparation, clean-up, and an almost unrelated code comment which would have avoided a lot of confusion on my part.
The pixel format conversion table would be good to share with
shm_buffer_get_cogl_pixel_format (), but where should I put it?