Upgraded min black to 26.3.1 and min python to 310

[Ritunjai-Sharma]

Associated Issue: #13

Github raised this security warning: https://gist.github.com/alon710/213feea7a4693b5774694763c9a88e80

To fix, had to upgrade black, which required upgrading min python to 310 (should be fine since we recently updated all repos to py 3.10)

Work done in this PR:

• Upgraded instances of black package in the repo to 26.3.1
• Upgraded instances of py 3.9 in the repo to 3.10. This means we have now upgraded min python to 3.10 in the repo
• Re-ran example notebooks in the repo using 3.10, since that's the new default python version rather than 3.9.
• The changelog had a typo whereby version 1.1.0 was mentioned twice. Have fixed that and also updated changelog with the changes made in this PR.

[cjmwills]

Changes all look good to me. One question is if/how we want to perform recon? Looking at the notebook diffs some of the importance values and order has changed. Assuming we don't want the situation where end users upgrading package version changes their outputs. Unless we're communicating that by the major version change?

[Ritunjai-Sharma]

Changes all look good to me. One question is if/how we want to perform recon? Looking at the notebook diffs some of the importance values and order has changed. Assuming we don't want the situation where end users upgrading package version changes their outputs. Unless we're communicating that by the major version change?

Good point! The changes in values and order are specific to example.ipynb and example_ohe.ipynb, which use XGBoost. These differences are likely due to variations in XGBoost, SHAP, or NumPy versions/compiled wheels between Python 3.9 and 3.10 environments. However, this release will indeed be a major update. Also, the existing doctest for XGBoost output should help identify any future changes.

The other notebooks that don't use XGBoost, i.e., binary_classification.ipynb, multiclass_classification.ipynb, and non_standard_model.ipynb, did not show changes in values or order. The only difference was in the number of decimal points displayed, which now aligns with the package's default precision of 16. This suggests the old outputs were outdated, and the new ones are correct.

[cjmwills]

Good point! The changes in values and order are specific to example.ipynb and example_ohe.ipynb, which use XGBoost. These differences are likely due to variations in XGBoost, SHAP, or NumPy versions/compiled wheels between Python 3.9 and 3.10 environments. However, this release will indeed be a major update. Also, the existing doctest for XGBoost output should help identify any future changes.

The other notebooks that don't use XGBoost, i.e., binary_classification.ipynb, multiclass_classification.ipynb, and non_standard_model.ipynb, did not show changes in values or order. The only difference was in the number of decimal points displayed, which now aligns with the package's default precision of 16. This suggests the old outputs were outdated, and the new ones are correct.

Makes sense thanks. Agreed the doctests help confirm no unexpected changes in that output too.