Med-dataloader

-------------------

* Added support one-hot encoding in case of multi-class label

------------------

* Added support for RGB Images
* Fixed some bugs related to norm_bounds types

------------------

* Main Changes in the package structure. Now there are two main functions:
generate_dataset and get_dataset, both leveraging on DataLoader class.
* The generation of the dataset can be handled also by CLI, to simplify usage.
* Processed data can live by themself. No more need to transfer also original
file (e.g. to Drive to make use of them on Colab)

------------------

* Improved flexibility for image data types. Now cache dimension reflects the
actual dataset dimension.

------------------

* Added support for 3D files: now Dataloader automatically detects whether a
file is 2D or 3D and returns the properly sized dataset. Please remember that
med_dataloader returns tf.data.Dataset object for 2D tasks, 3D is not yet
supported.
* Added new notebook in examples folder.

------------------

* Improved code flexibility:
* It is possibile to choose which type of data augmentation is performed
* Boundaries for data normalization can be set by the user
* Images can be resized automatically by the user
* Added basic_usage example also as a notebook