[PWGEM] Add ML-based photon cuts

[ikantak]

Add ML-based photon cuts to photonconversionbuilder, V0PhotonCut and Pi0EtaToGammaGamma(MC) in addition to normal standard cuts

[github-actions]

O2 linter results: ❌ 226 errors, ⚠️ 477 warnings, 🔕 0 disabled

[mhemmer-cern]

Looks good to me.

[ktf]

Hi, Notice this code uses KFParticle::GetQ which simply will not work on ARM, where char by default is unsigned. One needs alisw/KFParticle#9.

[ktf]

Now integrated in alisw/alidist#6086. Please make sure you validate the results if you are running this on ARM, because I simply did the mechanical substitution from char to signed char.

[mhemmer-cern]

Hi, Notice this code uses KFParticle::GetQ which simply will not work on ARM, where char by default is unsigned. One needs alisw/KFParticle#9.

Hey @ktf is there a way we could have known this before, when the person working on this does not have access to a Mac with ARM?
In any case thanks a lot for finding and quickly fixing this!

[ktf]

Actually Macs with ARM are fine, since there char is also signed by default. It can only be spotted on Linux on ARM and it was found by our code checker when building the O2 full system test on such platform. That said, I cannot guaranteed that everything is fine because if you do:

int charge = particle.GetQ();

you end up with positive charges only and you do not get any complains, so one would really need to check the details.

[mhemmer-cern]

I am confused, I thought with your PR GetQ() returns a signed char now, so should

int charge = particle.GetQ();

work without an issue now since on every system the return value for GetQ is ensured to be signed?

[ktf]

Yes, of course. What I meant is that there might have been other tasks where the old version was silently misbehaving. If that misbehaviour ended up in derived data, we would still be in trouble.

[mhemmer-cern]

Oh I see. Thanks a lot and lets hope that this is not the case!