use pixman for cursor plane rendering

[cmeissl]

update necessary after Smithay/smithay#1294 to use cursor plane for default cursor when rendering is needed

[YaLTeR]

Gave this a quick test, seems to work nicely, and render_cursor_plane in Tracy went from ~1.5 ms mean to ~0.4 ms mean, cool!

[cmeissl]

Gave this a quick test, seems to work nicely, and render_cursor_plane in Tracy went from ~1.5 ms mean to ~0.4 ms mean, cool!

This (and the comment in the smithay PR from drakulix about the format), reminded me that I forgot to check the
format in niri. So this now also includes the format correction. This has two nice effects, first the color (if you have colored
cursors) is now correct and it also enables the fast-path where we skip rendering all together if the size matches (no scaling or transform). So I would expect that if the cursor size matches the physical size it should take < 0.2 ms to copy the cursor.

[cmeissl]

Waiting for Smithay/smithay#1302

[YaLTeR]

Thanks! Yeah, now render_cursor_plane() has disappeared completely, save for three calls, for a similar test scenario to what I did above when it was in the hundreds. try_assign_cursor_plane(), whenever it is non-trivial, is in the 0.2-0.3 ms range (note linear time and log values):

There are just a few high outliers that seem to be some GPU stalls: