* refactor: numpy complex

* refactor: numpy float

* refactor: numpy bool

* refactor: numpy float64

test: fixes to bring tests up to date

* spark requirement + complex fix
* updated numpy object implementation
* old reference to bottleneck
* a regression in old numpy

resolves three bugs in the two types in the Python backend:
- EmailAddress had two relationships registered rather than a relationship and and a transformer
- URL contains_op had a typo
- URL relationship wasn't correctly checking for scheme and netloc.

* spark requirement + complex fix

* updated numpy object implementation

* old reference to bottleneck

* fixes complex tests and a regression in old numpy

* isort

* wip-setup

* removed pandas references in python tests;

* spark requirement + complex fix

* updated numpy object implementation

* old reference to bottleneck

* fixes complex tests and a regression in old numpy

* isort

* wip-setup

* removed pandas references in python tests;

* fix : import ABCIndex in place of ABCIndexClass

Pandas 1.3 renamed the ABCIndexClass to ABCIndex, so this should make the trick.
Backwards compatible.

Co-authored-by: Simon Brugman <sbrugman@users.noreply.github.com>