Current implementation of Geometry UDTs. PR commits still need to be squashed, but currently trying to solve a issue that will require some more development.

Currently all tests are passing locally, serialization + deserialization of UDT works in java RDDs and DataFrames. The current issue is a frontend display issue from notebooks with pyspark + pixiedust. We normally display the dataframe using pixiedusts display function i.e. display(df)
However, when trying to display with pixiedust the job fails because of how Spark tries to first convert the type to/from JSON before displaying. After more digging this will require a python UDT to be created in addition to the Java/Scala UDT. We should be able to use the Shapely library as a backing serialized object for each geometry type in python. This additional change may take some additional time just because some portions of how to work a python package into the current build process and deploy it properly with jupyter + spark are unclear to me.

Alternatively, We can use the builtin dataframe display
i.e. df.show()
Which will work and display the geometries in their WKT string format, but that makes our use of pixiedust relatively pointless in the notebooks beyond a job status tracker. I would ideally like the python package + UDT as the long-term solution. Going to attempt to get everything in place over the weekend.