true hyperbolic midpoints

As @dherdt pointed out, currently the geodesic midpoints are constructed using an Euclidean bi-section. This is to be replaced by a hyperbolic one.

Can this be done analytically?