requested to merge cory.quammen/vtk:pv-release-fix-multiblock-opacity into paraview/release
When deciding if a composite dataset has translucency, vtkCompositePolyDataMapper2 was incorrectly determining there was no translucency for a subtree of blocks if the visibility of the subtree root was set to false. This is incorrect because children could be set to visible, but the algorithm would not descend the tree to determine that.
Fixed this bug by moving the visibility determination down to just before the scalars translucency determination is made. If the block is not visibile, it does not have translucency.
Backport of !9768 (merged)