Option for simplified visibility cone generation purely from normals

[Jarkko Lempiainen]

Visibility cones are currently created by evaluating from which directions each meshlet is visible and fitting a cone per meshlet to these view directions. This accounts for mesh self-occlusion of rigid meshes and reduces the cone size thus enables more aggressive run-time meshlet culling. The standard and less optimal method is to fit triangle normals to the cone instead and we should add this method as well for comparison. This can be also useful to enable visibility cone culling for deformable meshes (e.g. skinned characters) since it's feasible to extend the cone angle at run-time based on deformation.