Work to expose spatial join directly through CLI. Updates algorithm to be run primarily using SpatialJoinRunner. There are only a couple areas of the code I don't feel entirely comfortable with:

1. The code surrounding the creation of the new data adapter for the join results particularly the naming of the new adapter. If there is a better way to come up with a unique adapter Id for the join that follows other areas where unique names are needed I would rather switch my code to follow that pattern.
2. The UDF registry class for spark udf functions. I hacked a bit for the WKT functions which is fine in my mind since that will be removed eventually, but the part that bothers me is a rather small switch but the change from 'g' to 'G' for the registered udf name in Spark. Technically this is a complete API change even though it is super tiny so it will break any older versions that were using those UDF functions. Should we keep maintaining those previous names? I chose not to in this PR because it adds a bunch of unnecessary code, but I can add them back in if we should keep supporting them.