backends/drm: offload atomic commits to a separate thread
This separate thread submits commits as late as possible, so that until immediately before vblank the cursor position can still be updated, reducing its latency and opening the door for more optimizations.
This doesn't wait on buffers being ready yet, so there can still be situations where the cursor lags. Fixing that is the next step.
As another possible optimization, with this thread we could remove the need for the scene to be involved in cursor updates, which should further improve responsiveness and efficiency.
Edited by Xaver Hugl