Adding FHMRUVV N3LO splitting functions

[giacomomagni]

This PR is imported from: https://github.com/NNPDF/eko_fhmv_private/pull/1

[felixhekhorn]

mostly minor stuff, but

• please check explicitly that you didn't break the LHA benchmarks since they were broken before your last commit and I'm not sure that fixes it (as it was numba complaining)
• we should remember to point them (=FHMRUVV) explicitly to this PR and tell them that they can (and should) start playing on their own

[giacomomagni]

Do we want to have to have use_fhmruvv also in the QED kernels?

[felixhekhorn]

Do we want to have to have use_fhmruvv also in the QED kernels?

is the effort manageable? if it is too much work, I'd say this is surely no priority ... if it is doable, we have to explicitly mention it when writing the mail ...

[giacomomagni]

is the effort manageable? if it is too much work, I'd say this is surely no priority ... if it is doable, we have to explicitly mention it when writing the mail ...

It's just adding few if statements, I can do it now.

EDIT: done in 517ae10

[felixhekhorn]

The moment https://github.com/NNPDF/eko/actions/runs/7489901631/job/20387536993 finished and you checked it, we can merge

[felixhekhorn]

Numba failed in https://github.com/NNPDF/eko/actions/runs/7489901631 ... but locally running poe lha -m "nnlo and sv and ffns" is just fine ...

I should have almost the same python (3.10.13 (me) vs. 3.10.10 (github)), the same numba (0.57.1), and the same numpy (1.24.4); ocard hash is the same (8beae75) (though theory not, but this was already discussed)

any ideas? @alecandido maybe?

[giacomomagni]

Numba failed in https://github.com/NNPDF/eko/actions/runs/7489901631 ... but locally running poe lha -m "nnlo and sv and ffns" is just fine ...

I should have almost the same python (3.10.13 (me) vs. 3.10.10 (github)), the same numba (0.57.1), and the same numpy (1.24.4); ocard hash is the same (8beae75) (though theory not, but this was already discussed)

any ideas? @alecandido maybe?

It might be due to banana!!
before n3lo_ad_variations had only 4 elements, now 7 can be required!

[giacomomagni]

Okay LHA benchmarks are now working again. 12918f9
Although the proper fix would be a new banana version...

[felixhekhorn]

Okay LHA benchmarks are now working again. 12918f9 Although the proper fix would be a new banana version...

good catch! let's merge if https://github.com/NNPDF/eko/actions/runs/7500258647/job/20418576973 goes through so we can move on