Re-compute polarized DIS predictions with fixed Yadism and EKO

[Radonirinaunimi]

We need to re-compute all the Polarized DIS FK table predictions making sure that:

• the grids include the updated type of initial state for the metadata
• they are generated with the fixed version of EKOs
• we no longer rely on the k-factors to compute $g_1$ and the asymmetries (because of improvements in n3fit)

These predictions include all the combinations of ZM and FONLL for the perturbative orders NLO and NNLO.

### Tasks
- [ ] https://github.com/NNPDF/yadism/pull/289
- [ ] https://github.com/NNPDF/yadism/pull/232
- [ ] https://github.com/NNPDF/nnpdf/pull/2122
- [ ] https://github.com/NNPDF/pinecards/pull/181

[giacomomagni]

The various tasks have been addressed in the corresponding PRs