Fix has_atomic_support check in can_use_hash_groupby()

[jbrennan333]

Closes #10583.

Change the has_atomic_support check in can_use_hash_groupby() to check the target type for the aggregation instead of the source type.

See discussion in #10583.

I have verified that this fixes the performance regression in our customer queries, and all unit tests still pass.

[codecov]

Codecov Report

Merging #10588 (423c216) into branch-22.06 (291fbcf) will decrease coverage by 0.00%.
The diff coverage is 96.07%.

@@               Coverage Diff                @@
##           branch-22.06   #10588      +/-   ##
================================================
- Coverage         86.31%   86.31%   -0.01%     
================================================
  Files               140      140              
  Lines             22300    22255      -45     
================================================
- Hits              19249    19210      -39     
+ Misses             3051     3045       -6     

Impacted Files	Coverage Δ
python/cudf/cudf/core/column/numerical.py	96.17% <ø> (ø)
python/cudf/cudf/testing/testing.py	81.69% <ø> (-2.82%)	⬇️
python/cudf/cudf/core/index.py	92.31% <90.90%> (+0.05%)	⬆️
python/cudf/cudf/core/single_column_frame.py	96.45% <93.75%> (-0.40%)	⬇️
python/cudf/cudf/core/column/categorical.py	89.77% <100.00%> (ø)
python/cudf/cudf/core/column/column.py	89.34% <100.00%> (+0.03%)	⬆️
python/cudf/cudf/core/column/datetime.py	89.71% <100.00%> (+0.39%)	⬆️
python/cudf/cudf/core/column/numerical_base.py	98.90% <100.00%> (+0.01%)	⬆️
python/cudf/cudf/core/column/string.py	89.10% <100.00%> (+0.12%)	⬆️
python/cudf/cudf/core/column/struct.py	96.42% <100.00%> (-0.09%)	⬇️
... and 10 more

---

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 090f6b8...423c216. Read the comment docs.

[jrhemstad]

Nice, I hadn't even thought of this as a solution. Very clean.

[PointKernel]

@gpucibot merge

[jbrennan333]

Thanks @jrhemstad, @PointKernel, and @ttnghia for the reviews, and @jlowe for help debugging this problem.