Set sample weight based on hash of row or key_column?

[samedii]

Currently we do it based on dataset index (dataframe index) and this is not nice when you want to communicate between datastreams