Draft: Don't limit DMA buffer screen casting only to Intel
Quoting the two commits:
screen-cast/src: Only use DMA buffers when explicitly asked for PipeWire clients specifies as part of the negotiation which buffer formats it wants to see. If not set to anything, `SPA_ID_INVALID` is passed, which is documented to mean "shouldn't need more than an mmap()". This is not true for DMA buffers, thus we should not automatically enable it for those. For those that explicitly asks for DMA buffers we shouldn't deny them if we have the ability to allocate them.
backend/native: Only disable dmabuf screen cast when not hw accelerated Don't attempt to predict whether we support DMA buffer screen casting depending on EGL context type or driver name, but only whether we are hardware accelerated. For all other cases, simply try and fall back to shared memory if need be. The reason we can do this now is that we only try to allocate DMA buffers if the PipeWire client asked for it, and rely on that client to know it can handle them. However, in case we're not hardware accelerated, we never want to attempt to use DMA buffer screen sharing, as we want to avoid compositing into a DMA buffer on such hardware as doing so can be very slow.
Related things that this relies on: https://gitlab.freedesktop.org/pipewire/pipewire/-/merge_requests/1067 and gnome-remote-desktop!63 (merged).