Update bot reference

[scarrazza]

We should include 2 validphys reports:

• with latest master as reference
• with the original state of the PR branch

[scarlehoff]

We should also change the runcard to leave only datasets that are included in NNPDF4.0. I'll update the code with the new reference so you can just copy the new results.

[Zaharid]

Does this issue still make sense?

[scarlehoff]

The idea of being able to do a bot PR against "master before PR" would still be good tbh but maybe more trouble than it is worth.