Versioning manifests in impresso_commons

[piconti]

We have devised a data versioning approach using manifests, as presented here.

Implement the code for such a manifest, which can be instantiated form the various processes:

• [x] Manifest Class
• [x] Data Statistics class
• [x] Ensure specific statistics computed correspond to each processing step
• [x] Document and add docstring on pycommons
• [x] Add user how-to and usage examples for NLP tasks

[piconti]

Fist implementation of the logic is done and has been sandbox tested at a small scale.

The next steps are now to add docstrings and document the interaction with the DataManifest instances, as well as to test it at larger scale, in particular with the upcoming patches that will be done to the canonical data.

[piconti]

The documentation of the manifests is in progress.

In addition to the original task, a few modifications and/or features need to be implemented:

• [ ] Adding the laste modification date at the year level
• [ ] Accelerate the aggregation process (aggregate by title to prevent very large dask graphs)