Pass nl as an argument in qed beta functions

[niclaurenti]

Implement APFEL running of alpha_em: below $m_{\tau}$, $\alpha$ runs with $n_l=2$.
Since EKO is never (or at least almost never) used below 1 GeV we don't care about switching off muon and electron running, nor setting $n_f=0$ below 0.5 GeV, as APFEL does.

[felixhekhorn]

please also adjust the documentation everywhere

[niclaurenti]

@felixhekhorn @giacomomagni regarding the conversation about the VFNS in the lepton number and the matching:
the reason why I am doing this PR is that this week I talked with SF and turned out that the photon I obtained was different from the one in NNPDF3.1QED since the $\alpha_{qed}$ running was different.
He wanted this running to be as close as possible to the one coming from the package alphaQED (http://www-com.physik.hu-berlin.de/~fjeger/software.html), that we should take as God-given truth.
Implementing the $\alpha_{qed}$ running as it is done in APFEL seems to mimic this running very well: this was probably the reason why they switched off $n_f$ at a certain point (moreover $\alpha_{qed}$ must exhibit a plateau for very small $Q$ and the only way to do it is to send the beta function to zero, i.e. $n_f=n_l=0$).
This is why I implemented this running first of all in the photon module in validphys and then here in EKO (in order to be consistent). I attach the plots to show what I'm saying: in the first one there is the wrong running, while in the second the running has been fixed
alphaQED23_vs_pineline4.pdf
alphaQED23_vs_pineline_mq05.pdf

In the end a test in which $m_q$ (the scale at which we pass to $n_f=0$) is varied
alphaQED23_vs_pineline_mq1.pdf

P.S. here in EKO, as I mentioned in the code, I only implemented the switch from $n_l=3$ to $n_l=2$ since we never (or almost never) go below 1 GeV, but in validphys I implemented all the thresholds since FiatLux requires $\alpha$ at very low scales.

[niclaurenti]

I have performed some benchmarks and the agreement with APFEL is the same of the master branch (as it should be since it is a tiny effect)

[felixhekhorn]

This will be in conflict with #335 as both touch QED kernels (should be simple to solve - however, but be aware)

[niclaurenti]

This will be in conflict with #335 as both touch QED kernels (should be simple to solve - however, but be aware)

There was no conflict

[felixhekhorn]

There was no conflict

even better (because you are touching sv.exponentiated.gamma_variation_qed and #335 ad_us.gamma_singlet_qed)

[felixhekhorn]

let's wait for https://github.com/NNPDF/eko/actions/runs/7501124078 and then merge