Add list_length scalar function

[mhk197]

Adds a list_length scalar function returning the number of elements in each list of a List array.

• Computed purely from the list's offsets/sizes — it never reads elements. Different paths for List and ListView arrays.
• Returns a U64 array; a null list yields a null length.
• Registered as a built-in (vortex.list.length) alongside list_contains, and exposed via the list_length(expr) expression constructor.

[codspeed-hq]

Merging this PR will improve performance by 13.97%

⚠️ 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.

⚡ 13 improved benchmarks
❌ 2 regressed benchmarks
✅ 1566 untouched benchmarks
🆕 6 new benchmarks

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Simulation	chunked_varbinview_canonical_into[(100, 100)]	223.8 µs	258.7 µs	-13.46%
❌	Simulation	chunked_varbinview_opt_canonical_into[(100, 100)]	304.7 µs	338.7 µs	-10.04%
⚡	Simulation	decompress_rd[f64, (10000, 0.01)]	139 µs	108.8 µs	+27.79%
⚡	Simulation	decompress_rd[f64, (10000, 0.1)]	139.3 µs	109.1 µs	+27.73%
⚡	Simulation	decompress_rd[f64, (10000, 0.0)]	138.9 µs	108.8 µs	+27.68%
⚡	Simulation	take_10k_first_chunk_only	271.8 µs	226.8 µs	+19.86%
⚡	Simulation	chunked_bool_canonical_into[(1000, 10)]	33 µs	27.6 µs	+19.56%
⚡	Simulation	take_10k_dispersed	285.5 µs	240.7 µs	+18.62%
⚡	Simulation	decompress_rd[f32, (100000, 0.0)]	583.6 µs	495.9 µs	+17.68%
⚡	Simulation	decompress_rd[f32, (10000, 0.1)]	91 µs	78 µs	+16.67%
⚡	Simulation	decompress_rd[f32, (10000, 0.01)]	90.8 µs	78 µs	+16.35%
⚡	Simulation	decompress_rd[f32, (10000, 0.0)]	91 µs	78.5 µs	+15.96%
⚡	Simulation	patched_take_10k_adversarial	260.4 µs	230 µs	+13.2%
⚡	Simulation	patched_take_10k_first_chunk_only	303.5 µs	273.1 µs	+11.14%
⚡	Simulation	patched_take_10k_dispersed	317.6 µs	287.2 µs	+10.59%
🆕	Simulation	list_large	N/A	9.9 ms	N/A
🆕	Simulation	list_medium	N/A	143.4 µs	N/A
🆕	Simulation	list_small	N/A	58.1 µs	N/A
🆕	Simulation	listview_large	N/A	6 ms	N/A
🆕	Simulation	listview_medium	N/A	98.2 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/list-length (0cdbea4) with develop (ed69077)}

[gatesn]

I think we use criterion these days right?

[mhk197]

I think its divan in vortex-array crate? Can switch to criterion though if wanted

[robert3005]

We only use criterion for cuda benchmarks, divan for the rest. Divan seems largely unmaintained, might need to build our own :/

[gatesn]

May as well support FixedList as well, then implement reduce to collapse it into the constant

[gatesn]

I think the "proper" way to do this would be to define an AnyList matcher, then execute_until::<AnyList>?