Panelsplit

New version release of panelsplit! Here are some of the changes:

- **More flexible**. The ability to work with numpy arrays for X and y. PanelSplit's period argument now accepts pd.Index, pd.Series, or np.array.

- **More modularized**. Now a lot of the PanelSplit functionality is split among different modules (e.g. cross_val_fit_predict is no longer a method of the PanelSplit class, but instead lives as its own function within panelsplit.application)
- **A pipeline class for panel data** There is a a new class, SequentialCVPipeline. This allows for the creation of pipelines which can be used in a panel data context.
- **Improved testing**.
- **Improved documentation**, which can be found [here](https://4freye.github.io/panelsplit/panelsplit.html).

This release fixes a bug that occurred when providing a Pipeline object as the estimator when using cross_val_fit. Now Pipeline objects can be provided as the estimator.

Cloning of transformer during _cross_val_fit ensures no leakage between splits.
Edited example in the examples folder after bug fixes.

Fixed the following issues:
https://github.com/4Freye/panelsplit/issues/15
https://github.com/4Freye/panelsplit/issues/13

These pertain to the fact that there was leakage in the estimator between splits. Now this is solved as the package clones the estimator before fitting to each split.

Fixed issues regarding the prediction function, namely https://github.com/4Freye/panelsplit/issues/7.

Initial release of panelsplit.