fix sum decimal array overflow behavior with scalars

[connortsui20]

Fixes #6429

Also refactors the code a bit so that we dont need a clippy allow complexity

[codspeed-hq]

Merging this PR will degrade performance by 10.22%

⚡ 5 improved benchmarks
❌ 2 regressed benchmarks
✅ 1128 untouched benchmarks
⏩ 1268 skipped benchmarks¹

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Simulation	encode_varbin[(1000, 2)]	203.4 µs	226.5 µs	-10.22%
❌	Simulation	encode_varbin[(1000, 4)]	205.3 µs	228.5 µs	-10.14%
⚡	Simulation	take_map[(0.05, 0.1)]	792 µs	603 µs	+31.34%
⚡	Simulation	take_map[(0.1, 0.1)]	1,118.8 µs	921.3 µs	+21.44%
⚡	Simulation	take_map[(0.05, 1.0)]	4.1 ms	3.1 ms	+33.35%
⚡	Simulation	take_map[(0.1, 1.0)]	4.7 ms	3.5 ms	+34.59%
⚡	Simulation	take_map[(0.1, 0.5)]	2.8 ms	2.1 ms	+34.07%

---

_{Comparing ct/fix-decimal-overflow-scalar (838d354) with develop (aa55054)²}

Footnotes

1. 1268 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

2. No successful run was found on develop (92156f5) during the generation of this report, so aa55054 was used instead as the comparison base. There might be some changes unrelated to this pull request in this report. ↩

[robert3005]

What was the bug here? That we returned overflow value even though it didn't fit into the declared dtype?

[joseph-isaacs]

overflow the precision but not the storage type?

[connortsui20]

That we returned overflow value even though it didn't fit into the declared dtype?

Yep, so the summation of the decimal array might not have fit the return dtype (even if it did not overflow the storage value). The regression test included in this change does something similar to what the fuzzer found.

[connortsui20]

overflow the precision but not the storage type?

Previously it only checked overflow on the storage type (in the sum_decimal and sum_decimal_with_validity functions with CheckedAdd::checked_add), now it does both

[robert3005]

Thanks, I should have read the test before asking my question