migrate tests from nbs to pytest

[deven367]

This PR tries to finish the first step in removing nbdev from the library. Tests from the core module have not been converted to pytest yet.

Edit: The tests from the core module have also been converted.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[CLAassistant]

All committers have signed the CLA.

[deven367]

@elephaint I can't reproduce this error locally
https://github.com/Nixtla/hierarchicalforecast/actions/runs/16532109297/job/46759619570

[JQGoh]

@elephaint I can't reproduce this error locally

@deven367 I can reproduce this locally using my mac platform. Will post here if I have new findings

[JQGoh]

@elephaint I can't reproduce this error locally

@deven367 I can reproduce this locally using my mac platform. Will post here if I have new findings

@deven367
Could you checkout and cherry-pick these two commits:

1. JQGoh@a4a8ecb
2. JQGoh@6f2fc44
   and include them into your branch? The first commit should help to resolve the MacOS's failed test case.

The main reason is that the UtW computed can differ slightly for Mac OSvs Ubuntu platform and this results in the instability for np.linalg.solve(), despite using the same numpy version 2.2.6

1. MacOS

array([[-5.68434189e-14,  0.00000000e+00,  2.84217094e-14,
         0.00000000e+00, -3.55271368e-15,  0.00000000e+00,
        -7.10542736e-15],
       [-1.42108547e-14,  0.00000000e+00,  7.10542736e-15,
        -8.88178420e-16, -1.77635684e-15,  0.00000000e+00,
        -3.55271368e-15],
       [ 0.00000000e+00,  0.00000000e+00,  2.84217094e-14,
        -3.55271368e-15, -7.10542736e-15,  0.00000000e+00,
        -1.42108547e-14]])

2. Ubuntu

array([[-2.84217094e-14,  0.00000000e+00,  1.42108547e-14,
        -1.77635684e-15, -3.55271368e-15,  0.00000000e+00,
        -7.10542736e-15],
       [-1.42108547e-14,  0.00000000e+00,  7.10542736e-15,
        -8.88178420e-16, -1.77635684e-15,  0.00000000e+00,
        -3.55271368e-15],
       [ 0.00000000e+00,  0.00000000e+00,  1.42108547e-14,
        -1.77635684e-15, -3.55271368e-15,  0.00000000e+00,
        -7.10542736e-15]])

[deven367]

1. JQGoh@a4a8ecb
2. JQGoh@6f2fc44
   and include them into your branch? The first commit should help to resolve the MacOS's failed test case.

Checking out the commits did fix the CI issues

[JQGoh]

Overall LGTM, left some comments to consider for improvements and a few questions to clarify. After your revision, I suggest that we can invite @elephaint to have a second look.

[elephaint]

Great work!

Few (minor) comments:

• I think wherever possible we should use Narwhals, avoiding the need for any pandas vs polars control flow
• Wherever .reconcile is called, S argument should be S_df to avoid getting deprecation messages.

[deven367]

• Wherever .reconcile is called, S argument should be S_df to avoid getting deprecation messages.

@elephaint It's fixed now!

[deven367]

The recent release of polars to 1.32.1 breaks the CI because of

>       raise NotImplementedError(msg)
E       NotImplementedError: `pivot` is only available in 'polars>=1', found version '0.50.0'.

Hence the version is now being pinned to <=1.32.0
cc @elephaint

[elephaint]

Great work @deven367, thanks!

Also @JQGoh thanks for helping