Investigate scaling workflow.

[YooSunYoung]

The scaling workflow fails and it's probably because there are too many nan values after processing.
We would like to go through step by step.

Here is the documentation of the overview of data processing:
https://scipp.github.io/essnmx/generated/modules/ess.nmx.scaling.average_roughly_scaled_intensities.html

[YooSunYoung]

One of the example dataset @aaronfinke gave me makes these scale factors:
(I don't know how it's supposed to look like though)

[aaronfinke]

That looks ok actually (scaling will depend on degree of chebyshev polynomial). That is not what I was getting in my workflow, however. I don't know what I am doing wrong.

[YooSunYoung]

@aaronfinke
I tried including different binning of wavelength and it affects the result quite a lot.
This plot above is made with a whole range of wavelength, 250 bins.

[jokasimr]

Do we have a minimum reproducible example where the scaling fails? It would be very useful to debug the procedure.

[aaronfinke]

Don't worry about scaling for now. Fixing it will require a much more thorough deep dive that I have not gotten around to yet.