Disallow groupby aggs for StructColumns

[charlesbluca]

Closes #8474

We were erroneously using the _STRING_AGGS set of allowed aggregations for struct dtypes in groupby.pyx, which allowed us to perform erroneous groupbys on StructColumns; in @ayushdg's example:

df = cudf.DataFrame(
    {
        'a':['aa','aa','cc'],
        'd':[{"b": '1', "c": "one"}, {"b": '2', "c": "two"}, {"b": '3', "c": "one"}]
     }
)

df
	a	d
0	aa	{'b': '1', 'c': 'one'}
1	aa	{'b': '2', 'c': 'two'}
2	cc	{'b': '3', 'c': 'one'}

df.groupby('a').collect()

	d
a	
aa	[{'0': '1', '1': 'one'}, {'0': '2', '1': 'two'}]
cc	[{'0': '3', '1': 'one'}]

This change corrects this error, which should now prevent groupby operations on StructColumns.

[codecov]

Codecov Report

❗ No coverage uploaded for pull request base (branch-21.08@0a4e8a1). Click here to learn what that means.
The diff coverage is n/a.

❗ Current head e31f9a4 differs from pull request most recent head bfa53fb. Consider uploading reports for the commit bfa53fb to get more accurate results

@@               Coverage Diff               @@
##             branch-21.08    #8499   +/-   ##
===============================================
  Coverage                ?   82.91%           
===============================================
  Files                   ?      110           
  Lines                   ?    18094           
  Branches                ?        0           
===============================================
  Hits                    ?    15002           
  Misses                  ?     3092           
  Partials                ?        0           

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 0a4e8a1...bfa53fb. Read the comment docs.

[marlenezw]

Nice catch Charles, this looks good to me.

[quasiben]

Thanks @charlesbluca

[quasiben]

@gpucibot merge

[quasiben]

@charlesbluca do you want to add a test for this as well ?

[charlesbluca]

Sure! I'll make sure we catch a pandas.core.base.DataError in the case of struct column groupbys.