Add list_sum expression

[mhk197]

Adds a vortex.list.sum scalar function: the sum of each row's list elements, one output row per input row — the equivalent of DuckDB's list_sum() and DataFusion's array_sum().

Note that list entries that are empty or all null need to be nullified. However, GroupedAccumulator returns 0 in these cases, so we need to either:

1. Change the behavior of GroupedAccumulator and grouped sum (or even sum) kernels or
2. Do a second pass of list element validity and sizes, construct a mask for the sums, and apply it.

The first option is probably best but can be applied in a followup PR.

[codspeed-hq]

Merging this PR will degrade performance by 18.28%

⚡ 1 improved benchmark
❌ 2 regressed benchmarks
✅ 1616 untouched benchmarks
🆕 11 new benchmarks
⏩ 42 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Simulation	copy_nullable[65536]	573.2 µs	1,027.7 µs	-44.23%
❌	Simulation	copy_non_nullable[65536]	794.2 µs	908.6 µs	-12.59%
⚡	Simulation	bitwise_not_vortex_buffer_mut[128]	273.6 ns	244.4 ns	+11.93%
🆕	Simulation	fsl_sum_large	N/A	65.2 ms	N/A
🆕	Simulation	fsl_sum_medium	N/A	727.8 µs	N/A
🆕	Simulation	fsl_sum_small	N/A	56.5 µs	N/A
🆕	Simulation	list_sum_large	N/A	135.7 ms	N/A
🆕	Simulation	list_sum_medium	N/A	1.1 ms	N/A
🆕	Simulation	list_sum_nullable_elements_large	N/A	495.9 ms	N/A
🆕	Simulation	list_sum_nullable_elements_medium	N/A	4.9 ms	N/A
🆕	Simulation	list_sum_small	N/A	143.4 µs	N/A
🆕	Simulation	listview_sum_large	N/A	108.1 ms	N/A
🆕	Simulation	listview_sum_medium	N/A	899.6 µs	N/A
🆕	Simulation	listview_sum_small	N/A	73.4 µs	N/A

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing mk/list-sum-expr (f3354dd) with develop (24ad8fd)}

Footnotes

1. 42 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. ↩

[myrrc]

Let's push list_sum down to Duckdb and write a sqllogictest for this (can be in a followup PR). I'd also like to see a performance benchmark query, I'll prepare the machinery for that.

[myrrc]

I think we want AnyColumnar here and handle constant/canonical version at once

[robert3005]

Ah, this is handled above in execute

[robert3005]

ok no, the code above will be forced to canonicalise

[mhk197]

ah yeah we want to canonicalize a 1 row constant array, do sum, and then broadcast. will fix