Implement a fix_convolution method

[Radonirinaunimi]

The following PR allows fixing one of the convolutions and return a new Grid with lower dimension. For eg. it can be used to convert a SIDIS grid into DIS.

TODO:

• simplify a bit the logic
• improve the error handling and CLI interface
• perform extensive checks (checked in DY-like and single-inclusive hadron production cases)
• propagate into the APIs

cc @vbertone

[Radonirinaunimi]

With this, I believe this is now done.

[cschwan]

I'm not sure the following situation is covered: one PDF + one FF and you integrate out the PDF, thereby getting rid of the factorization scale (logs). Can you check that?

[Radonirinaunimi]

I'm not sure the following situation is covered: one PDF + one FF and you integrate out the PDF, thereby getting rid of the factorization scale (logs). Can you check that?

I implemented that specific case in 3775166 and e26d360.

[cschwan]

This is almost perfect, can you please add the following tests to up the code coverage: call Grid::fix_convolution with

1. a DIS grid to trigger the first error and
2. an invalid index to trigger the second error.

Ideally we'd also test missing the missing coverage where DIS+FF-grid's first convolution is fixed, but I can imagine creating such a grid isn't easy.

[Radonirinaunimi]

This is almost perfect, can you please add the following tests to up the code coverage: call Grid::fix_convolution with

1. a DIS grid to trigger the first error and

2. an invalid index to trigger the second error.

Added in 2faed55.

Ideally we'd also test missing the missing coverage where DIS+FF-grid's first convolution is fixed, but I can imagine creating such a grid isn't easy.

Also done in 2faed55. I didn't add it initially because I was lazy xP but actually the Python tests ecosystem allows for such implementations to be done relatively easy.