Issue/401/use qp+fix zinteg

[m-aguena]

Using qp for handling PDFs. The main changes are:

• The interpolation of the PDFs on a finer grid for integration in _integ_pzfuncs is done by qp.
• Adding the possibility of having quantiles as the PDF information:
  • Added this option in generate_galaxy_catalog
  • Makes the GCData object convert the quantiles into a PDF grid with GCData.get_pzpdfs, which is used in the integration mentioned above.

Closes #401 and #643

[coveralls]

coverage: 100.0%. remained the same
when pulling be75a04 on issue/401/use_qp+fix_zinteg
into 7d90ee1 on main.

[hsinfan1996]

1. Should we add qp to requirements.txt and the Requirements section in README?

2. It seems to me that the imports in some of the files are not compliant with PEP8, but since pylint doesn't complain about C0411 / wrong-import-order, it is not a big deal. I am just wondering.

[combet]

This new notebook is very useful to see the various behaviors, but the markdown text needs to be corrected/expanded to help the user understand better what is happening. At the moment, the markdown seems to copied/pasted from another demo notebook and gives little information as to what is actually happening.

[combet]

Would be good to add a legend in the figure comparing the pdz computed for the 3 types (individual, shared and quantiles)

[combet]

Also, maybe add some explanation text for the last figure of the notebook. I find trick to understand what is actually shown.

[combet]

And why is the qp integral performing worse than the clmm-implemented ones but for the "quantiles" case where qp appears to perform better?

[m-aguena]

Thanks for the comments! I added more explanation to the notebook

[m-aguena]

I think that qp is performing worse than clmm because it kind of used a type like integral with the precision set by the separation of the zbins of the input provided, and in clmm we are refining the grid for the integral.

In the case of using quantiles, I think qp might be outperforming clmm because it is actually using the statistical meaning of quantiles while clmm recreates an interpolated PDF and integrates it, thus loosing information.

[combet]

Thank you @m-aguena. It's great to see qp and the quantile option available!

I've tested the code and everything appears to be working well. I just have comments about the new notebook showing the integral with qp versus clmm, etc. At the moment, I don't think they are enough or clear enough markdown explanations to easily understand what this is doing.

(And sorry, I clicked on "comment-only" for some comments and on "review" so they don't all appear in the "review" below)

[combet]

Also, maybe add some explanation text for the last figure of the notebook. I find trick to understand what is actually shown.

[combet]

And why is the qp integral performing worse than the clmm-implemented ones but for the "quantiles" case where qp appears to perform better?

[hsinfan1996]

I added qp (qp-prob) to both requirements.txt and environment.yml.

[combet]

Thanks for adding more explanations to the notebook. All looks good to me know!