ENH: Implement `spatial_shuffle` about dask-geopandas

[jorisvandenbossche]

Note that this is basically calling dask's set_index with the calculated distance:

ddf_shuffled = ddf.set_index(ddf.geometry.hilbert_distance())

But, one thing that we probably also want during that process is to calculate the spatial bounds of each partition, so that the resulting ddf_shuffled.spatial_partitions is set.
I am not fully sure if we can do that more efficiently by tightly integrating that into the set_index (which could mean copying some of the dask implementation to be able to adapt it), or whether calling a ddf.calculate_spatial_partitions() afterwards is as good.

[martinfleis]

I don't think we can make it any faster than calling one after the other.

[jorisvandenbossche]

One question from the meeting: we could add a keyword to control whether to actually add the calculated distance as the index or to drop it.

[jorisvandenbossche]

This might be worth having a separate issue about, but one of the ideas mentioned in a previous meeting is to also have a way to shuffle two dataframes at the same time with a consistent partitioning (or shuffle one dataframe to align with the partitioning of another).