Phenonaut

Added
- PackagedDataset for JUMP-MOA compound set, allowing easy access to this metadata
- added percent_compact performance measure
- functionality to metrics.utils for manipulating percent compact results into plots
- added percent_replicating_results_dataframe_to_95pct_confidence_interval to phenonaut.metrics.utils, allowing the calculation of confidence intervals on percent replicating scores by resampling the null distribution.
- added restrict_evaluation_query to phenonaut.metrics.performance.percent_replicating to allow restriction of evaluation to certain rows without removing unused compounds from the pool of non-matching compounds comprising the null distribution.

Changed
- Renamed phenonaut.metrics.utils.percent_replicating_results_dataframe_to_percentile_vs_replicating to percent_replicating_results_dataframe_to_percentile_vs_percent_replicating and changed returned fraction replicating to percent replicating (X100) to keep terminology consistent.
- Changed the above newly renamed percent_replicating_results_dataframe_to_percentile_vs_percent_replicating return_fraction (by default True) to return_counts (by default False). If True, then y values denote the counts of replicates which were deemed replicating. If False, then percent replicating is returned.
- Renamed phenotypic metrics with magic values in phenonaut.metrics.non_ds_phenotypic_metrics from metric_with_magic_values to metrics_with_magic_values

Added
- added NaN handling capability within phenonaut dataset class
- added Tests for NaN handling functionality

Added
- revert method to Phenonaut object, enabling quick reload of already saved Phenonaut objects.
- arguments and capabilities to percent_replicating, enabling: automatic naming of performance files from parameters, saving run parameters to a JSON file.
- added new percent_replicating_summarise_results function to metrics.utils, allowing interrogation of files or directories containing percent_replicating results.
- phenonaut.metrics.non_ds_phenotypic_metrics containing non phenonaut Dataset aware generic similarity/distance methods.

Fixed
- fixed bug in Phenonaut.save, whereby overwrite_existing had no effect, and existing files were always overwritten.

Added
- metrics.utils with function for manipulating percent replicating results into plots
- phenonaut.metrics.performance.percent_replicating can now operate with other similarity and distance functions.

Fixed
- phenonaut.metrics.performance.percent_replicating now uses Spearman's rank as per the Cell paper (Way, Gregory P., et al. "Morphology and gene expression profiling provide complementary information for mapping cell state." Cell systems 13.11 (2022): 911-923).

Fixed
- fixed issue concerning continuing to work on Phenonaut objects after saving
- percent replicating

Changed
- remove_blocklist_features now considers prefixed columns and allows removal of features and non-feature columns.
- moved percent replicating to a new submodule: phenonaut.metrics.performance

Added
- get_df_features_perturbation_column function to Phenonaut and Dataset objects allowing clean interface for code to get at underlying DataFrames regardless if passed a Phenonaut object or Dataset.

Fixed
- fixed missing install requirements (torch, torchvision, torchaudio)