Update get_cardinality_distribution to support NaNs in child_column

[gsheni]

Problem Description

• Currently, the get_cardinality_distribution does not take into account NaN foreign keys (in the child column).
• To support NaNs foreign keys in SDMetrics, we would like to update this function to take this situation into account.

Expected behavior

• If there are NaNs values in the child column, they should be accounted for in the get_cardinality_distribution
• Change child_column.value_counts(dropna=True), taking into account NaNs
• If a child foreign key is null, that means that the parent is also null.
• Add tests to test_utils.py

Example

• Let's say we have the following:

  parent_column = pd.Series([1, 2, 3, 4, 5])
  child_column = pd.Series([3, 4, 1, 4, 4, 5, 1, np.nan, np.nan])

• The distribution would be as follows:

  parent  child_counts
  0       1           2.0
  1       2           0.0
  2       3           1.0
  3       4           3.0
  4       5           1.0
  5       np.nan      2.0