Probably wrong thermal noise level for cross-pols

[steven-murray]

We're seeing significantly low thermal noise on cross-pols in the HERA validation data. It might be that the cross-pol noise is getting assigned using a single autocorrelation instead of the geometric mean of both autos.