Normalization due to windowing
It was recently found out that the normalization applied to data in
LALInferenceReadData was not consistently applied on templates for f-domain models. This should affect us too, though the correction factor should be absorbed in the wavelet amplitude. Still we should add this factor explicitly, rather than be careful every time we are trying to interpret the amplitudes. Once the fix has been merged in LI, apply it to BW (or BWP) too.