[BUG] XGBoostLSS dynamic dependencies: add optuna to python_dependencies when n_trials != 0

[ravjot07]

Summary

This PR fixes an issue where XGBoostLSS requires optuna when n_trials != 0 for hyperparameter optimization, but the dependency was not properly declared in the python_dependencies tag.

Depends on #681 for testing.

Changes

• Modified XGBoostLSS.__init__: Added logic to dynamically include optuna in python_dependencies when n_trials != 0

Problem

When n_trials is set to a non-zero value, XGBoostLSS uses hyperparameter optimization via the _hyper_opt method, which requires optuna. However, the python_dependencies tag only included xgboostlss, causing dependency checking to fail when optuna was not installed.

Solution

The fix dynamically updates the python_dependencies tag at the end of __init__:

• When n_trials == 0: Only xgboostlss is required (no hyperparameter optimization)
• When n_trials != 0: Both xgboostlss and optuna are required (hyperparameter optimization enabled)

This ensures the dependency checking system properly validates and installs optuna when needed, without removing test coverage.

[fkiraly]

Thanks!

Kindly simplify, see above.

[fkiraly]

Thanks!

[fkiraly]

the failure comes from an oversight in VM testing that does not take into account that more dependencies may need to be installed than in the class tag. I have to think about how to resolve this, but it is not an issue for your PR.