Spin phase angles

This MR implements a different spin convention. This uses the lalsimulation conversion between the spherical basis and the cartesian spins.

In this case it is more natural to sample in phi_12, the difference in azimuthal spin angles, and phi_jl, an azimuthal spin angle relative to the total angular momentum.

This will be basically invisible and no-one will ever want to change this prior (bold statement).

@gregory.ashton @paul-lasky is this considered a friendly change?