Additional metrics for conservation / continuity / divergence

[ma595]

To discuss with Dominic https://arxiv.org/abs/2510.13722
the headline technique in this paper is I think a Power Spectral Density, which they then use in a loss function.