Uncut python predictions shape doesn't match ndata

Created by: wilsonmr

Example with NMC:

>>> from validphys.api import API
>>> from validphys.fkparser import load_fktable
>>> from validphys.convolution import central_predictions, predictions
>>> ds = API.dataset(dataset_input={"dataset":"NMC"}, use_cuts="nocuts", theoryid=53)
>>> cuts = ds.cuts.load() if ds.cuts is not None else None
>>> fks = [load_fktable(fk).with_cuts(cuts) for fk in ds.fkspecs]
>>> fks[0].ndata
292
>>> pdf = API.pdf(pdf="NNPDF31_nnlo_as_0118")
>>> predictions(ds, pdf).shape, central_predictions(ds, pdf).shape
((234, 100), (234, 1))

for some reason the convolutions are giving 5 extra points, I'm not sure to what extent this applies to other datasets. It's not caught by any tests because we never test NMC predictions without cuts.