Don't frustum cull shadow casting objects
Previously in ShadowMapPass::renderPrep, shadow casting objects were
collected but they were culled if they were outside of the camera
frustum view. This change makes it so that no objects are culled and so
that shadow casting objects are always part of the shadowmap.
For performance reasons all the object selection and bounds calculations
are done in a new function QSSGLayerRenderData::getShadowCastingObjects.
This avoids creating extra lists and sorting them.
Fixes: QTBUG-145014
Pick-to: 6.11
Change-Id: Id3e86a36d6b083713daee64506e611b8e1449e47
Reviewed-by: Christian Strømme <christian.stromme@qt.io>
Reviewed-by: Andrei Yankovich <EndrIIMail@gmail.com>