Fix remaining luminosity combinations

[cschwan]

It seems I wasn't thorough enough in #25 and missed a few lines. I noticed that while implementing NNPDF/pineappl#199.

I can't guarantee the first commit fixes everything, so I suggest we keep this open until I've tested it extensively.

[scarlehoff]

Is this ready to be merged? (I forgot about it)

[cschwan]

I'm no longer sure, I forgot about it too. From my comment it seems it isn't ready yet. I'll have look at it later.

[cschwan]

As far as it concerns me, we can now merge this. I fixed all bugs that I discovered using symmetry arguments, but this really is untested, and we should devise a way to test this (in a different Issue though). At the very least we should study the impact of these changes.

[scarlehoff]

Ok, I'm going to merge this. We probably want to define some bunch of checks in order to make sure that everything works as expected. The problem is that I'm not sure there is any other code with isoscalar target (without hacking) we can compare against?

[cschwan]

A check that we could perform is that we

1. generate a proton-proton grid,
2. rewrite the initial state of one proton to the one of an iso-scalar, meaning that we perform the replacements u -> (u + d)/2 and d -> (u + d)/2 (also for the corresponding anti-particles), which will require support in PineAPPL and
3. compare that against the same calculation with collider = piso.