Crossmatch different ZTF data releases

[bfhealy]

Light curve are now available for ZTF DR13. The golden dataset's classifications were done using ZTF DR5, and that release's ZTF source ids (_id column in Kowalski) are not necessarily the same as those in DR13. We will have to find or make a crossmatch between these releases if we want to use the latest features to train our model.

(Potentially a separate issue is whether the existing labels should change based on the new data. It is possible the golden classifications are of high enough confidence that the additional data will not change them.)

[AshishMahabal]

If the golden set classifications are high confidence, they will most likely not change. But a test at some point will be useful. That is why we carry confidances along.