v0.2.3

Previously, output partitions were labelled sequentially as part.{i}.parquet for i in 0...n, where n was the number of unique H3 parent cells in the dataset, one part per parent.

To make this easier to consume, we now label these as {h3parent}.parquet, which allows users to select a particular partition based on a H3 cell without having to do anything except list the output directory. The output order is the same since the sort order was already based on a lexicographic order of the H3 cells.