Treatment of zero bins with zero errors from HEPData
Some HEPData entries (see for example table 17 of this record: https://www.hepdata.net/record/ins1499471 ) have zero entries with zero uncertainty. This is interpreted in some circumstances as an infinitely precise measurement of zero. This happens in contur, which is where I spotted it, but I can imagine it being an issue for tuning/fitting or other applications as well. There are ways of working around it (which I'll implement for now in contur) but it would be good to have a general solution. IMO these are incorrect entries uploaded to HEPData by the experiments. @GraemeWatt, @20DM what do you think?