Investigate to_parquet by_layer=False

[dougbrn]

There's an issue with the parquet files generated by to_parquet when by_layer=False

Copied from code comment:
Divisions cannot be generated from a parquet file that stores nested information without a reset_index().set_index() loop. It seems like this happens at the to_parquet level rather than in read_parquet as dropping the nested columns from the dataframe to save does enable divisions to be found, but removing the nested columns from the set of columns to load does not. Divisions are going to be crucial, and so I think it's best to not support this until this is resolved.