Mapping order on amdgpu

[emersion]

Somewhat related to https://github.com/emersion/libliftoff/pull/60.

Trying to present these 3 layers on amdgpu:

1. Fullscreen NV12
2. Fullscreen ARGB8888
3. Cursor ARGB8888

This combination can work:

1. Primary plane
2. Overlay plane
3. Cursor plane

However, libliftoff will try to construct the mapping step by step:

• Will try to map layer (1) to the primary plane, success
• Will try to map layer (3) to the cursor plane, failure because amdgpu can't do cursors directly over YUV

It's a case where adding a new plane will make the configuration work -- the intermediate ARGB plane allows amdgpu to draw the cursor.