Visual Localization in 3D Maps: Comparing Point Cloud, Mesh, and NeRF Representations