Implement flavor masks

[alecandido]

• [ ] boolean mask function: given an FkTable, you return a boolean mask (you can implement as a method: FkTable.flavor_mask(), no argument)
  • we should also find in which way flavors are sorted inside vp: for the boolean mask changes nothing (we have a mask consistent for the order used by vp), but in order to use it, we should correctly sort our PDF first
• [ ] convolution function: given an FkTable and PDF, return the convolution (you can also make a method FkTable.convolute(pdf))
• [ ] extend for hadronic datasets

The last point is a bit more subtle, because it is more convenient to have a PDF as input (uniform FkTable.convolute signature with DIS, and at the end a PDF fully determines the input) but you need to mask luminosity, so:

• we need to find out how luminosities are sorted (I hope in a sensible and stupid way (i, j) -> i*14 + j or something similar)
• we need to virtually construct the external product array $v v^T$: virtually to make an efficient operation (make a kind of a view)

This last part we can discuss together @tgiani.

[alecandido]

@tgiani I turned your commit into a PR, such that we can discuss here, and try to solve separately the problem.

Notebooks are hard to review (have a look to the "Files changed" tab and you find why by yourself ;D). So it is fine to use them to experiment, but try to turn them as part of the library as soon as possible. The further benefit is that directly developing a library makes everything more maintainable: it is easier to reuse and refactor.