Scalecast

Added
- added `backcast()` method to `Forecaster` and `MVForecaster` objects
- added `copy()` and `deepcopy()` methods to the same objects
- added `test_only` options to all models in `Forecaster`. ignored in combo models. a future dist will have this for `MVForecaster`
- added `plot_loss_test` as arg in lstm and rnn models
- added `notebook.results_vis_mv()` function
Changed
Fixed

Added
Changed
- changed how names are looked for in the `MVForecaster` object
Fixed

Added
Changed
Fixed
- fixed an issue when plotting in MVForecaster that occurs when you didn't initially pass names to `__init__()`
- fixed an instance where labels was spelled wrong

Added
Changed
Fixed
- Fixed inconsistent indexing in MVForecaster that was causing some models to fail when different Forecaster objects' dates didn't align
- Fixed an issue where `level=True` wasn't working in `plot_test_set()` sometimes

Added
- Added more description to the `MVForecaster.__repr__()` method
Changed
- Changed arguments in the rnn estimator to make it more flexible and dynamic
Fixed

Added
Changed
Fixed
- Noticed an error when forecasting with Silverkite on differenced data and fixed it.