Introduce a programmatic way for producing urban sprawl result

[delhomer]

As the operations are nicely organized from overpass requests to result plotting, we may undertake a data pipeline formalization, for instance through a Python ETL like Luigi (see doc).

The idea would be to ensure that each operation is well done, and that there is no need to redo it. This should be done through a CLI based on Luigi capabilities, let imagine the following command to plot any urban sprawl metric:

luigi --module urbansprawl.tasks PlotWhateverMetric --city-ref lyon --date-query 2018-12-20 ...

[delhomer]

After PR #3 one may retrieve OSM data and compute urbansprawl indices; after PR #6, one may extract INSEE population features (French use case) and prepare data for further supervised algorithm executions.