Make a partitioning based on the clustering of the data

[JulienPeloton]

Idea 1 (fixed clustering):

• load raw data
• perform a k-means where k = number of partitions
• repartition accordingly

Idea 2 (dynamic clustering):

• Load raw data
• Look dynamically for clusters in the data. Maybe we can start with some guess, and increase/decrease if data suggest.
• repartition accordingly