Dataset support other partitions

[mahiki]

Dataset currently defines a rundate partition in a hive-style layout. This can be handy for s3 queries with Athena or Spark, or other partitioned file reader. Dataset is getting in the way of read/write across partitions, or writing a large dataframe partitioned by columns.

Not sure how important this is, usually the pipeline is generating small datasets at the end of a long process, and Dataset is there just to store these artifacts in an organized way rather than scattered all over my laptop.

Probably Prefect will be publishing their Julia API before this becomes a problem.