Add sum_v2 with SQL empty-input semantics

[mhk197]

Summary

• add SumV2 with the distinct vortex.sum_v2 ID and an explicit { sum, is_overflow, is_empty } partial
• return null for empty or all-null input and for integer or decimal overflow, while preserving valid zero sums
• register a specialized grouped primitive kernel and route pushed DuckDB SUM through SumV2
• keep legacy Sum, Stat::Sum, file statistics, zoned compatibility, Mean, and list_sum unchanged
• update DuckDB SQL logic expectations for all-null sums

This implements the follow-up direction from #9113 after its revert in #9324. It introduces a new aggregate ID instead of changing the serialized contract of vortex.sum. It follows the compatibility direction explored in #8705, using the requested sum_v2 name and migrating the DuckDB pushed aggregate.

[codspeed-hq]

Merging this PR will degrade performance by 4.12%

⚠️ Unknown Walltime execution environment detected

Using the Walltime instrument on standard Hosted Runners will lead to inconsistent data.

For the most accurate results, we recommend using CodSpeed Macro Runners: bare-metal machines fine-tuned for performance measurement consistency.

⚡ 3 improved benchmarks
❌ 4 regressed benchmarks
✅ 1974 untouched benchmarks
🆕 15 new benchmarks
⏩ 54 skipped benchmarks¹
🗄️ 4 archived benchmarks run²

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Simulation	sum_i32_nulls_scattered	483.8 µs	647.5 µs	-25.28%
❌	Simulation	canonicalize_sparse_list[(512, 7, 4)]	571.7 µs	654.9 µs	-12.7%
❌	Simulation	canonicalize_sparse_list[(1024, 17, 8)]	508.7 µs	577.5 µs	-11.92%
❌	Simulation	cold_misaligned[(16, 64)]	380.2 µs	431.3 µs	-11.84%
⚡	Simulation	sum_i32_nullable_all_valid	93.8 µs	79.6 µs	+17.81%
⚡	WallTime	words_gather_scalar_avx2[65536]	9.4 µs	8.3 µs	+13.47%
⚡	Simulation	take[small_m/shuffled/primitive/nonnull/chunks=2048/indices=16]	376 µs	341.8 µs	+10.03%
🆕	Simulation	canonical_sum_v2_f64_all_valid	N/A	243.1 µs	N/A
🆕	Simulation	canonical_sum_v2_f64_clustered_nulls	N/A	288.7 µs	N/A
🆕	Simulation	canonical_sum_v2_i32_clustered_nulls	N/A	278.1 µs	N/A
🆕	Simulation	canonical_sum_v2_i32_nullable_all_valid	N/A	246.1 µs	N/A
🆕	Simulation	sum_v2_f64_all_valid	N/A	226.2 µs	N/A
🆕	Simulation	sum_v2_f64_clustered_nulls	N/A	272.5 µs	N/A
🆕	Simulation	sum_v2_i32_clustered_nulls	N/A	261.9 µs	N/A
🆕	Simulation	sum_v2_i32_nullable_all_valid	N/A	220.1 µs	N/A
🆕	Simulation	sum_v2_f64	N/A	233 µs	N/A
🆕	Simulation	sum_v2_f64_nulls_clustered	N/A	246.3 µs	N/A
🆕	Simulation	sum_v2_i32	N/A	101.8 µs	N/A
🆕	Simulation	sum_v2_i32_nulls_clustered	N/A	122.8 µs	N/A
🆕	Simulation	sum_v2_i32_nulls_scattered	N/A	486.5 µs	N/A
...	...	...	...	...	...

ℹ️ Only the first 20 benchmarks are displayed. Go to the app to view all benchmarks.

Tip

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

---

_{Comparing mk/sum-v2 (9c95149) with develop (e4b3421)}

Footnotes

1. 54 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. 4 benchmarks were run, but are now archived. If they were deleted in another branch, consider rebasing to remove them from the report. Instead if they were added back, click here to restore them. ↩

[myrrc]

LGTM for duckdb changes