Save write options for training datasets

[SirOibaf]

We provide a way for users to pass write options to spark when writing a training dataset. We should store these options in Hopsworks and retrieve them when reading/appending to the training dataset.