The source data is not accessible if the user gets the code via a dbc download

Our current dbc export does not include arbitrary files, only notebooks.

It would be ideal if we download the source data from Github to dbfs before reading the data, so that the accelerator can be run from the repo and the dbc.