Wrong conversion from boost::geometry::point to spherical_coslat_differential

[div5252]

While conversion to spherical_coslat_differential, the longitude term should have been multiplied with cos(latitude) but it isn't so.
That is we get the same result as spherical_differential, which is incorrect.