Fix closure_results to account for data keyword
Created by: wilsonmr
Now you can actually use the closure estimators for a single closure fit
I've just changed the collects here and made sure experiments
key isn't in the report
Here is an old comp fits I ran ages and ages ago: https://vp.nnpdf.science/fweyXKx9RQWbBcofOgF2cw==/
Here I check the numbers are reproduced: https://vp.nnpdf.science/BtNgsoMESXWdf6oeJHhiqw==/
I'm satisfied that the numbers are reproduced. The discrepancies are small and in functions which bootstrap and at any rate are consistent up to bootstrap error.