Scalecast

Added
- `util.pdr_load()` now accepts multiple series and returns an MVForecaster object of everything loaded together
- added more arguments to the `util.pdr_load()` function
- added `auxmodels.vecm` model, which is a model class that can be imported using the `MVForecaster.add_sklearn_estimator()` function (11)
- modified the source code in the `MVForecaster` object to accomodate new model classes (e.g. vecm)
- added a vecm grid to the example grids
- added the `util.find_optimal_coint_rank()` and `util.find_optimal_lag_order()` functions
Changed
- changed scaling syntax in `Forecaster` and `MVForecaster` to circumvent a warning having to do with feature names--only numpy arrays are scaled now (not dataframes)
Fixed
- added a call of `Forecaster.Forecaster.typ_set()` right after `MVForecater.__init__()`, before chopping dates to fix weird loading errors that occured once in a while
- fixed the util function that wasn't working in 0.14.2 and yanked that release. everything scheduled for 0.14.3 will now be part of 0.14.4.

Added
- if there are not enough observations to use in cross validation (usually because too many AR terms were added), an error is raised when calling the `Forecater.cross_validate()` and `MVForecaster.cross_validation()` functions
Changed
- no `Forecaster.auto_Xvar_select()` no longer raises errors if more AR terms passed to max_ar argument than the model is able to estimate
Fixed
- `Forecaster.determine_best_series_length()` will no longer fail if the min_obs arg value is greater than the amount of observations in the series
- found more instances where `TypeError`s should not be raised (such as passing an `int64` type when `int` is required)
- fixed an issue that occurs after selecting Xvars with `Forecaster.auto_Xvar_select()` on an integrated series then loading to `MVForecaster`
- fixed the error raised when 0 or less is passed to the `Forecaster.set_validation_length()` and `MVForecaster.set_validation_length()` functions

Added
- added the `Forecaster.auto_Xvar_select()` method
- added a check for NAs in `Forecaster` and `MVForecaster` when evaluating grids and validation metric is mape. a descriptive error is raised if NAs are found (10)
- added `Forecaster.drop_all_Xvars()`
- added `Forecaster.determine_best_series_length()`
- added `Forecaster.restore_series_length()`
Changed
- removed deprecated functions identified and labeled in 0.13.1
- `Forecaster.keep_smaller_history()` can now accept numpy int types as an argument.
Fixed

Added
- added grids_file attribute to `Forecaster` and `MVForecaster` objects, as well as `set_grids_file()` method to both objects.
Changed
Fixed

Added
- added probabilistic argument option to `auxmodels.mlp_stack()` function.
- added Xvars argument to `auxmodels.auto_arima()` function.
Changed
- made it so that an error is raised earlier when using `Forecaster.ingest_Xvars_df()` incorrectly.
Fixed
- fixed an issue with `auxmodels.mlp_stack()` where `**kwargs` were not being passed correctly.

Added
- added error arg to `Forecaster.tune_test_forecast()` and `MVForecater.tune_test_forecast()` methods.
Changed
- took out an error check that was redundant and not even working
Fixed
- fixed some documentation syntax for new objects added last dist