operations should check for duplicated columns

[SirOibaf]

It's easy when joining feature groups together to end up with duplicated columns. Spark refuses to write if there are duplicated columns in the dataframe.

When we check the schema we should also check that there are no duplicated columns.