Scalecast

Added
Changed
Fixed
- fixed exception catching in `Forecaster.auto_Xvar_select()` and `util.find_optimal_transformation()`

Added
- added `dayofyear` to search for in `Forecater.auto_Xvar_select()` when frequency is daily or lower.
- added `cycle_lens` argument to `Forecaster.add_seasonal_regressors()`.
Changed
Fixed
- fixed how seasonalities are selected when they are not Fourier transformed in `Forecaster.auto_Xvar_select()`.

Added
Changed
Fixed
- fixed getting cis with backtesting. only the first backtest iteration was being used.

Added
- added `method` argument to `Forecaster.reeval_cis()` and `MVForecaster.reeval_cis()` and included an option to get confidence intervals through backtesting.
- added `Forcaster.plot_backtest_values()`.
Changed
- changed some error messages to be more descriptive and to encourage raising issues on github.
- changed how level confidence intervals are obtained when calling `SeriesTransformer.DiffRevert()` to be more efficient.
- the backtest_values DataFrame now includes dates in `Forecaster` and `MVForecaster`. the order of the dataframe is Date --> Actuals --> Preds for all iterations.
Fixed

Added
Changed
- changed `dynamic_testing = <int>` to same way it was in 0.12.8 when it was introduced because it gives better and more efficient results in both `Forecaster` and `MVForecaster`.
Fixed

Added
Changed
- the `m` argument in `util.find_optimal_transformation` can be a list and multiple seasonal differences can be tried in that function
- changed how level plotting is peformed in `Forecaster.plot()` and `Forecaster.plot_test_set()`
Fixed
- fixed how history['LevelY'] attribute is undifferenced in `SeriesTransformer`, which was causing plots to be incorrect.