Update the reweight tool for z expension model

[LiangLiu212]

• Add reweighting support for the z-expansion axial form factor.

• Update the reweighting implementation for the z-expansion vector form factor.

Weights are computed by scaling the cross section using standard error propagation.

• Disable the single-parameter dial tweak mode, in this ways, not support for varying parameters according to Cholesky decomposition to get the standard deviation. I don't this it is useful for error estimation. But I can re-implement it if we need this.

[nusense]

Can you check the question about using kXSecTwkDial_ZExpELFF where I'd expect kXSecTwkDial_ZExpZAFF.

I also checked with Jeremy Wolcott to see whether NOvA is using the individual ZExpELFF tweak parameters. We agreed that these seem to be vector form factor parameters and NOvA doesn't use those, only the axial, so I don't think we're pulling the rug out from under them.

[nusense]

might as well claim the copyright through the year it's getting committed: 2026. Here and elsewhere.

[LiangLiu212]

I will bump copyright.

[LiangLiu212]

Can you check the question about using kXSecTwkDial_ZExpELFF where I'd expect kXSecTwkDial_ZExpZAFF.

I also checked with Jeremy Wolcott to see whether NOvA is using the individual ZExpELFF tweak parameters. We agreed that these seem to be vector form factor parameters and NOvA doesn't use those, only the axial, so I don't think we're pulling the rug out from under them.

As I understand, the previous implementation of reweight of axial form factor can only include the correlation matrix by using grwghtnp and we need to feed covariance matrix when we run the reweight.

However, I think we should consider the uncertainties of the ZExp model as a whole, rather than focusing on the individual parameters. And we don't need to have a feature to tweak individual parameter.

So, in my PR, I provide a new method to estimate the uncertainties of ZExp model with covariance matrix by using the error propagation method.

[LiangLiu212]

I just realized that we might still want to reweight the Z-expansion from one set of parameters to another.
However, I think this should be treated as a separate feature rather than part of the uncertainty estimation. I will think about how to implement it.