fix(array): honor the execution allocator for RowFn outputs

[connortsui20]

Depends on #10016.

RowFn output payloads bypass the execution allocator. This change allocates them directly through ctx.allocator() across owned, deferred-retry, selected, filtered, constant, and sink execution, while preserving zero-copy primitive publication and empty-output paths.

OutputElement chooses its collection storage through an associated buffer type and an allocation hook. The executor writes through OutputBuffer slots, and the buffer implementation constructs the array. Vortex primitive and Boolean implementations use BufferMut, while scalar and fixed-size-list sinks use the same storage contract. UTF-8 descriptors, external bytes, and polygon payloads also use the execution allocator. Physical sink parameters remain separate from allocation resources.

Regressions check ownership of returned payloads using canonical inputs prepared before allocation tracking, including a context override, constant UTF-8 output, retry execution, and zero-copy reuse. A zero-sized output with Vec storage exercises the owned execution paths without requiring BufferMut. Boolean collector selection is preserved.

The Boolean dense-retry path from merged #9986 also uses the execution allocator, with coverage in the existing packed-output allocator test.

Validation before the rebase: 168 focused comparison, mask, and RowFn tests passed on the combined stack through #9979, along with cargo clippy -p vortex-array --all-targets --all-features -- -D warnings. Tests, formatting, and benchmarks were not rerun locally after the rebase.

[codspeed]

Merging this PR will degrade performance by 1.39%

⚠️ 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 benchmarks measured no execution time

Nothing ran under measurement, usually because the compiler removed the code under test. These results are not comparable, so they count as unchanged.

Preventing compiler optimizations

⚠️ Different runtime environments detected

Some benchmarks with significant performance changes were compared across different runtime environments,
which may affect the accuracy of the results.

Open the report in CodSpeed to investigate

⚡ 2 improved benchmarks
❌ 5 regressed benchmarks
✅ 2173 untouched benchmarks
⏩ 385 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	WallTime	filtered_owned_i64_avx2[OneNullInEight]	21.8 µs	26 µs	-16.07%
❌	WallTime	decode_avx512[8192, (Inline, OneNullInEight)]	80.9 µs	95.1 µs	-14.97%
❌	WallTime	decode_avx2[8192, (External, OneNullInEight)]	152.3 µs	174.3 µs	-12.63%
❌	WallTime	dict_canonicalize_gt_u8_neon[1000000]	492.3 µs	548 µs	-10.16%
❌	Simulation	non_nullable[256]	196.3 µs	218.2 µs	-10.06%
⚡	Simulation	decode_primitives[f32, (1000, 32)]	35.9 µs	22 µs	+63.58%
⚡	WallTime	decode_avx512[8192, (Inline, AllValid)]	112.8 µs	102.5 µs	+10.03%
⚠️	Simulation	fixed_16_advancing_ptr_safe[100]	< 1 ns	< 1 ns	N/A
⚠️	Simulation	preverify_advancing_ptr_unchecked[1000]	< 1 ns	< 1 ns	N/A
⚠️	Simulation	preverify_advancing_ptr_unchecked[10000]	< 1 ns	< 1 ns	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 ct/row-fn-output-allocator (8e3b4c0) with develop (eff2b1e)}

Footnotes

1. 385 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. ↩

[connortsui20]

(blocked on #10030)

[gatesn]

Nit: rename allocate to with_capacity to align with Vec etc?