DOC : Add example for the loading and pre-processing of corpus

[Thibeb]

Example for loading and pre-processing the QUAERO, E3C and CASM2 corpus

Each corpus is first converted to a medkit text document format using a custom function.
Then each text document is processed into a pipeline to tokenize sentences and split the doc into multiple mini docs for better futur training.

The corpus_specs function takes a list of medkit text documents and analyse it to get some metrics such as the number of original docs, sentences, entities and their repartition.