Add realistic test case for convolute_eko

[felixhekhorn]

in order to not rely on a manual test (e.g. in pineko) we should add a proper test here, for that we need:

• a simple grid G: this can be a positivity-like grid with, e.g., O1 = [u(x=0.1, Q2=100)] with a sufficent, but not too big number of bins in x
• an EKO E: the "raw" eko should be sufficient to keep the actual library out of the game, say E(10 <- 100)
• a PDF set f: to make the grids/table collapse to numbers

the actual check would then be f(Q0) x (E x G) = f(Q)

this could also be reused for #108

PS: corresponding email thread around Wed, Mar 02, 2022 at 11:27:26AM +0100

[cschwan]

@andreab1997 @alecandido @felixhekhorn I'm now working on this and using NNPDF/pineko#62 as a test case. By the way, the crate ndarray_npy is very handy to load the EKO.

[alecandido]

By the way, the crate ndarray_npy is very handy to load the EKO.

Good to know, it is one of those we considered for the Rust loader NNPDF/eko#97