Cut over sum aggregate function

[gatesn]

Move sum compute function over to the Sum impl of AggregatFnVTable

Note that we remove the ability to sum extension types since we were incorrectly just summing the storage DType. This doesn't make sense, e.g. summing timestamps is invalid, but summing time deltas is allowed.

[codspeed-hq]

Merging this PR will degrade performance by 10.2%

⚡ 6 improved benchmarks
❌ 1 regressed benchmark
✅ 1002 untouched benchmarks
⏩ 1515 skipped benchmarks¹

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

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
⚡	Simulation	take_map[(0.1, 1.0)]	3.8 ms	3.5 ms	+10.6%
⚡	Simulation	null_count_run_end[(10000, 256, 0.1)]	41.6 µs	37.4 µs	+11.07%
⚡	Simulation	null_count_run_end[(10000, 256, 0.5)]	41.7 µs	37.2 µs	+12.23%
⚡	Simulation	null_count_run_end[(10000, 4, 0.01)]	140.5 µs	113.7 µs	+23.56%
⚡	Simulation	null_count_run_end[(10000, 1024, 0.1)]	40 µs	35.7 µs	+12.18%
⚡	Simulation	null_count_run_end[(10000, 1024, 0.5)]	40.1 µs	35.5 µs	+12.79%
❌	Simulation	map_each[BufferMut<i32>, 128]	770.6 ns	858.1 ns	-10.2%

---

_{Comparing ngates/sum-agg (8ae7b12) with develop (da4c196)}

Footnotes

1. 1515 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. ↩

[robert3005]

We should make sure we port any exhisting behaviour we have in the current implementations

[gatesn]

Like what?

[joseph-isaacs]

NaN handlings at least. Its work fuzzing this on the side is possible

[robert3005]

I left a comment on the pr that added sum with the code location since that code isn’t touched in this pr. If we want to change decimal sum to always work in i256 we should make this in a separate pr

[robert3005]

FWIW NaNs are a long standing issue #5152. I think fixing NaN handling is what's actually required for this stat to be universally useful but it didn't change in this pr

[gatesn]

Added bonus, benchmarks are 10-15% faster!