Introduce per-phase sample-detector displacement with flat detector

[clemisch]

Introduces model for sample-detector displacement, assuming transmission geometry and a flat detector as presented in Hulbert2023.

Unlike other displacements (m2ThetaDisplacement, m2ThetaTransparency) which act only on the whole PowderPattern (i.e. all phases), this new displacement can act also per-phase. In each PowderPatternDiffraction it combines per-sample and per-phase displacement and computes resulting 2θ shift from it.

m2ThetaFlatDetDispRatio is the sample displacement, and m2ThetaPhaseFlatDetDispRatio is the per-phase displacement.

GSAS-II has something similar called LayerDisp here which allows per-phase offsets. Afaik it's currently undocumented.

[clemisch]

I did notice a wrinkle which originates from allowing displacement on both sample and phase and combining them inside the phase:

I have a non-zero m2ThetaFlatDetDispRatio on the sample, none on the phase. So I compute the reflections 2θ positions via PowderPattern.X2XCorr (manually for e.g. a plot), but the displacement m2ThetaFlatDetDispRatio is not taken into account and the reflections are wrong.

This is confusing but by design: we can't compute 2θ shifts from sample and phase displacement independently and then combine them, because they don't combine linearly. We have to compute the effective displacement sample+phase and then compute 2θ shifts. And that we can only do in PowderPatternDiffraction.X2XCorrPhase, i.e. per-phase.

For refinement this is no problem because it happens behind the scenes. But it makes manual computation like above more complicated. See e.g. https://gitlab.esrf.fr/cschmid/xrdmap/-/commit/a806b05049ff2931719babff0b6b8fcabf810cac.

I still want to have per-phase offsets, but I can see this confusion is a strong argument against it.