BUG: Fix pivot_table margins to include NaN groups when dropna=False

[iabhi4]

Fix incorrect margin computation in pivot_table when index or columns contain NA values

This PR fixes an issue where the "All" row or column (i.e., margins=True) in pd.pivot_table does not account for rows that contain NA values in the index or column dimensions. These rows were incorrectly excluded from the overall aggregation used to compute the margin, leading to incorrect totals.

The fix modifies the margin calculation to ensure that rows with NA values are included in the aggregation, consistent with how the data is treated in the main table when dropna=False.

• closes BUG: margin for pivot_table is incorrect with NA column/index values #61509
• Test added and passed
• Ran pre-commit check
• Added entry in doc/source/whatsnew/v3.0.0.rst under Reshaping

[iabhi4]

This change fixes the margin behavior in pivot_table when dropna=False, but as expected, it also affects related logic, a few tests are now failing, especially in test_crosstab, where the old margin behavior is still being asserted.

Before I go ahead and update those tests to match the new behavior, I just wanted to double-check if this is the direction we want to take, treating dropna=False more consistently in how margins are computed. If this looks good, I’ll go ahead and fix up the failing tests accordingly

[rhshadrach]

@iabhi4 - thanks for putting this up. The crosstab failures here look like bugs to me as well. E.g. in the penultimate column

b    one        two        NaN       All
c   dull shiny dull shiny dull shiny    
a                                       
bar    1     0    1     0    0     0   2
foo    2     0    1     1    0     1   5
All    3     0    2     1    0     0   7

All should be 1, as this PR is now producing.

[iabhi4]

@rhshadrach Thanks for confirming! I’ve updated the test cases to reflect the corrected behavior

[github-actions]

This pull request is stale because it has been open for thirty days with no activity. Please update and respond to this comment if you're still interested in working on this.

[rhshadrach]

Looks good, just a small request.

[rhshadrach]

Can you revert this change. The numbers in the third column remain floats because of np.nan. The change here makes it look like the result is object dtype.