PU tool update normalisation cut-off

For some reason the PU tool cuts off if the sum of weights is smaller than a hard-coded value. This causes severe problems if the cross section of the process is too small (see discussion on PAT-help: https://groups.cern.ch/group/hn-atlas-PATHelp/Lists/Archive/Flat.aspx?RootFolder=%2Fgroup%2Fhn%2Datlas%2DPATHelp%2FLists%2FArchive%2FHuge%20PU%20weights&FolderCTID=0x0120020084D7E80CB0C3394A8D54EF07C309B044) Since Will isn't around anymore I guess it's hard to know what the reason for this is, but if someone sees an issue feel free to comment.