Scalecast

Added
- Added an explicit error for when `auxmodels.mlp_stack()` is specified incorrectly.
- Added `ax` argument to more plotting functions.
- Added test scripts to the GitHub repository.
- Added an explicit error in `Forecaster.auto_Xvar_select()` for situations when monitoring a test metric but no test set is specified.
Changed
- Explicitly specify `dtype = float` for parts of code that raise a warning when this is not the case.
- Changed to relative library imports wherever possible.
- Changed the `Forecaster.__deepcopy__()` to be more explicit.
- Changed how `GridGenerator` gets grids and added more grids files to choose from.
- Got rid of some confidence-interval information when exporting model summary info from `MVForecaster`.
- `MVForecaster.export('lvl_test_set_predictions')` no longer fails when there is no test set.
Fixed
- Fixed the `AnomalyDetector.EstimatorDetect()` function.
- Fixed `util.break_mv_forecaster()` for situations when `MVForecaster` does not have a test set.
- Fixed how exporting works in `MVForecaster` so that errors are not returned when test set is 0 and default args are maintained.

Added
Changed
- Explicitly specify `dtype = float` for parts of code that raise a warning when this is not the case.
- Logging a warning from `SeriesTransformer.DetrendTranform()` instead of raising it.
Fixed

Added
- Added `FutureWarning`s to functions that touch the `Forecaster.diff()` method. This method will be removed soon as the `SeriesTransformer` object is a better alternative that does the same thing.
- Added `diffy` argument to `Forecaster.adf_test()`.
- Added a `correct_residuals()` function that corrects residuals for autocorrelation before building confidence intervals.
Changed
- The `train_only` default argument in `auxmodels.auto_arima()` changed to False.
- Changed how default arguments are parsed in `Forecaster.export()`.
- Refactored code to avoid a `SettingWithCopyWarning` from pandas in `Forecaster.ingest_Xvars_df()`.
Fixed
- Removed the `print_attr` argument from `notebook.results_vis()` that is no longer accepted in `Forecaster.diff()`.
- Removed a warning from `Forecaster.ingest_Xvars_df()` that shouldn't have been raised.

Added
- Added the `cis` argument to `Forecaster__init__()` and `MVForecaster.__init__()`.
Changed
- Only conformal intervals now supported in `Forecaster` and `MVForecaster`. By default, these will not be generated and can only be generated if there is a test set and it is specified to a sufficient length.
- Removed `Forecaster.proba_forecast()` and `MVForecaster.proba_forecast()` and all probabilistic arguments in functions.
- Default `test_length` argument in `Forecaster.__init__()` and `MVForecaster.__init__()` is now 0.
- Took out the `exog_coint` argument from `vecm.__init__()` since it's not actually usable.
- Only model-specific and hyperparameter optimization warnings will be logged in warnings.log and other warnings will be printed.
- Changed many categories of warnings from `UserWarning` to `Warning`.
Fixed
- Added `staticmethod` decorators to `util.metrics` methods. This doesn't change functionality, just the documentation.

Added
Changed
Fixed
- Fixed the warning in `Forecaster.ingest_Xvars_df()`.

Added
Changed
Fixed
- Fixed an error that can arise from the `util.break_mv_forecaster()` function.
- Fixed the `AnomalyDetector.EstimatorDetect()` function.