Join cardinality validation shall not raise on multiple null values (that never produce matches)

[sibbiii]

Description

According to the documentation, the validate parameter of polars.DataFrame.join checks whether
join keys are unique in the left/right/both datasets.

Consequently,

left = pl.DataFrame({'a': [1, 2, None, None]})  # unique key and some missing data
right = pl.DataFrame({'a': [1, 1, 2, 2]})  # non-unique key

_ =left.join(right, how="left", on="a", join_nulls=False, validate="1:m")  

fails with polars.exceptions.ComputeError: join keys did not fulfill 1:m validation.
but this is clearly a one to many join.

Note: nulls are not joined (join_nulls = False),
so the left dataframe is unique (except for the nulls).

Use case

Real world datasets often have null values for missing data (the example above is just greatly simplified). On such datasets, no cardinality validation can be performed on a left join.

I googled a bit and tied to find similar issues but could not find one. I am also not sure if my question here makes sense (the current behaviour is as documented) but in my option the example above is clearly a one-to-many join with some missing data.

Expected behaviour

Exclude null values from the uniqueness check if join_nulls=False (null values will not produce matches)
The example above shall therefore not raise a ComputeError.

(Note: I am not sure if the current behaviour is only because of Pandas.
original issue that added join cardinality validation: #9263)

[barak1412]

@orlp May you assign me?