Draft: qt: implement static optimization hint in kawase blur

As the comment says, this allows releasing the intermediate layers (if live is set to false) in four pass (default) configuration. It has no benefit in two pass configuration because it already uses only one layer, and we need that.

With nvidia-smi, I observed around 10 MiB reduction in video memory consumption in the player page when I tried a few album covers.

Hopefully this will help with #26908. I'm not sure if it should be closed though.