Correct F3 and g1 in NC at NNLO and N3LO

[giacomomagni]

We were using the wrong combination of CC coefficients to implement the non singlet NC F3 at NNLO and N3LO.
This might explain why Fig3 of the Yadism paper was much worst than the other benchmarks and might have an impact on the polarized g1

TODO:

• Check the impact on g1 (FONLL polarized paper)
• Check the benchmark with Apfel++ g1 (Yadism paper Fig3) @giacomomagni
• Check impact on some FKtables
• Check the impact on the IC valence paper, F3 plot. @giacomomagni

@felixhekhorn @Radonirinaunimi @RoyStegeman

[Radonirinaunimi]

Thanks for this @giacomomagni! I can help checking the 1st and 2nd point.

[giacomomagni]

Thanks for this @giacomomagni! I can help checking the 1st and 2nd point.

I'm taking care of point 2.

[Radonirinaunimi]

Thanks for this @giacomomagni! I can help checking the 1st and 2nd point.

I'm taking care of point 2.

Ok, I will only take care of 1 then.

[giacomomagni]

I couldn't spot any major difference. If you agree we can proceed to merge.

[Radonirinaunimi]

I have just compared the results with the pineappl grids used for the FONLL POL paper (although for ZM only) and the differences are numerical. So this indeed do not significant changes.