Pv3/Pv3HM waveform model bug fixes
@sylvain.marsat if you have a moment could you review these changes too?
This MR consists of two commits to fix two bugs:
commit 1: It also includes a small fix to help the model work in Bilby as was pointed out by @rory-smith. This fix simply shifts the zero frequency array element away from zero so that the precession angles are not evaluated at exactly zero which can sometimes cause NaNs to occur.
commit 2:
The final spin was being calculated only using chi1x it should have been using chip.
This was pointed out by the PhenomX review team specifically by @marta.colleoni
I have compared PhenomPv3 and PhenomPv3HM before and after the changes using standard waveform group regression test code and the results can be found below. Follow the links to find corner plots coloured by the mismatch.
To summarise the difference before and after the bug fix seems to be quite small. Out of 5000 random test cases the worst match was about 0.996 with the majority of cases have matches above 0.999.