Support ext-image-capture-source-v1 and ext-image-copy-capture-v1

[Ferdi265]

This completely flew under the radar for me. These two protocols were just merged into wayland-protocols and if implemented in wlroots would potentially allow window screenshare in wlroots compositors.

This has less priority than #16 since xdg-desktop-portal-wlr will likely expose window capture as well, but this protocol likely has better selection ergonomics on some desktops (allowing to specify an output on the commandline).

[Ferdi265]

wlroots has just merged ext-image-copy-capture-v1 and ext-image-capture-source-v1 (without toplevel capture for now).

Implementing this is also likely not as complicated as implementing xdg-desktop-portal handling (since no dbus and no pipewire is involved). I'll attempt to implement this in two stages: first for shm only and then for dmabuf, since dmabuf here means we need to create our own dmabufs (similar to screencopy with dmabuf) rather than just importing one (wlr-export-dmabuf and pipewire).

Implementing the toplevel capture source is a top priority once it lands in wlroots as it will finally allow window mirroring in wlroots compositors.

[Ferdi265]

Initial working example for output capture via ext-image-copy-capture-v1 in branch feature-ext-image-copy-capture.

This includes a WIP refactor of SHM buffer and DMABUF allocation, which will potentially also allow the screencopy backend to use DMA-BUF capture. If that happens, there will be another fallback option for non-DMA-BUF screencopy for setups where DMA-BUF doesn't work