Issues on PhenomD and PhenomPv2

[SSL32081]

This is a duplicate of tedwards2412#13.

Summary
There are some small issues concerning PhenomD and PhenomPv2:

• Use the 3rd order spline to estimate fRD (as is done in LAL), which could further reduce the mismatch between ripple and LAL.
  • I think keeping the autodiff in ripple is perhaps better, as it would be more faithful to the equations themselves. Though it is good to know where the source of error is.
• Discrepancy between PhenomD and PhenomPv2 in LALSuite.
• Discontinuity in lal's phenomPv2 right at in-plane spin = 0
  • This may have been addressed by Add aligned-spin conditions to PhenomPv2 angle calculations #3 ?

We will need to first reproduce these claims.

[ThibeauWouters]

I was re-reading Zipeng Wang's technical note on the implementation of Pv2 when he worked on this ~year ago. It seems he already investigated this and the mismatch got reduced indeed (not sure about the speed). Therefore, it might be a good idea to have an in-depth comparisons of the autodiff and spline interpolation again to see if it would be worth it to switch to splines.

[ThibeauWouters]

@thomasckng