[BUG] Python 3.10 regression fails

[joocer]

see results:
https://github.com/mabel-dev/opteryx/actions/runs/2287677990

Other versions do not fail

[joocer]

Appears to be related to NULL records not be deduplicated

[joocer]

This may be related to this change in Python 3.10

Hashes of NaN values of both [float](https://docs.python.org/3/library/functions.html#float) type and [decimal.Decimal](https://docs.python.org/3/library/decimal.html#decimal.Decimal) type now depend on object identity. Formerly, they always hashed to 0 even though NaN values are not equal to one another. This caused potentially quadratic runtime behavior due to excessive hash collisions when creating dictionaries and sets containing multiple NaNs. (Contributed by Raymond Hettinger in [bpo-43475](https://bugs.python.org/issue?@action=redirect&bpo=43475).)

https://docs.python.org/3/whatsnew/3.10.html

[joocer]

FIXED