Closes #1138
Closes #1141 [here](https://github.com/rapidsai/cuspatial/pull/1156/files#diff-c522c9afb3364b1aed2b2589c0d0c260dbc634bc54844536b1d382cecb92bf29R112)
Depends on #1152
Depends on #1064

Please direct your attention [to the notebook](https://github.com/rapidsai/cuspatial/pull/1156/files#diff-cc4c516f63efa822793d75315c1b28a04bad6c9efc6fd2bdcac5cc30b05d14dd) since the dependencies and delayed state of CI issues over this week have put a lot of files into this PR.

This notebook demonstrates cuSpatial's new binary predicates on large datasets, benchmarking and comparing against the host implementation on GeoPandas.

In order to support the large inputs for these comparisons I had to reactivate the `pairwise_point_in_polygon` functionality that I'd previously written off. This is because quadtree doesn't support large N for NxN operations, since it is many-to-many, and brute-force would require a huge number of iterations to support such large dataframes. There are some more optimizations that can be made to speed up `pairwise_point_in_polygon`, but the algorithm itself isn't sufficiently fast. It is detailed fairly closely in the notebook.

Please take a look and let's have some conversations about steps forward.

Authors:
  - H. Thomson Comer (https://github.com/thomcom)

Approvers:
  - Michael Wang (https://github.com/isVoid)
  - Mark Harris (https://github.com/harrism)
  - Ray Douglass (https://github.com/raydouglass)
  - AJ Schmidt (https://github.com/ajschmidt8)

URL: #1156