lir 1.6.0 on Python PyPI

What's Changed

Within a data_strategies, the splitting strategy can be automatically determined by selecting one of the auto_* strategies.
Added the leave-one-source-out strategy to data_strategies.
Function modules defined in YAML now support keyword arguments.
feature_data_csv: added a feature_columns argument to explicitly specify which columns should be interpreted as features.
feature_data_csv: Added an option to continue processing when encountering None values during reading.
Added a new module SelectInstances for selecting elements from an InstanceData object by their index number.
Added a --set command-line argument to override YAML values.

registry: renamed modules.logit to modules.logistic_calibrator.
Improved the default output generated by bootstrap.yaml.
The KDE calibrator now accepts misleading predictions with log-odds of inf and -inf.
Redesigned and repurposed csv_writer, and moved it to a dedicated module.
Replaced the lir.visualization module by a single aggregation class, which takes parameters from registry entries.
Removed obsolete module wrapper classes.

Fixed the CSV parser error message prefix.
Fixed issues on the experiment setup documentation page and corrected several docstrings.
Fixed various minor code and documentation issues.

Full Changelog: v1.5.1...v1.6.0