Implement native interleave for ListView

[vegarsti]

Closes #9342.

This PR adds a native implementation of interleave for the ListView type. Also adds a benchmark. Unfortunately this uncovered that the fallback implementation is broken, see the test_list_view_interleave_fallback_is_broken test, which panics if we comment out the native implementation:

---- interleave::tests::test_list_view_interleave_fallback_is_broken stdout ----

thread 'interleave::tests::test_list_view_interleave_fallback_is_broken' (17537938) panicked at arrow-data/src/equal/utils.rs:102:59:
range start index 16 out of range for slice of length 0

But with the native implementation, the benchmark results are comparable to those for list. I have not been able to make them faster than list for this benchmark.

 ┌───────────────────────────┬────────┬──────────┐
 │ Benchmark                 │ List   │ ListView │
 ├───────────────────────────┼────────┼──────────┤
 │ (0.1,0.1,20) 100          │ 2.88µs │ 3.04µs   │
 ├───────────────────────────┼────────┼──────────┤
 │ (0.1,0.1,20) 400          │ 10.1µs │ 10.7µs   │
 ├───────────────────────────┼────────┼──────────┤
 │ (0.1,0.1,20) 1024 3-slice │ 33.7µs │ 34.1µs   │
 ├───────────────────────────┼────────┼──────────┤
 │ (0.1,0.1,20) 1024 4-slice │ 30.1µs │ 33.9µs   │
 ├───────────────────────────┼────────┼──────────┤
 │ (0.0,0.0,20) 100          │ 1.41µs │ 1.58µs   │
 ├───────────────────────────┼────────┼──────────┤
 │ (0.0,0.0,20) 400          │ 9.51µs │ 11.7µs   │
 ├───────────────────────────┼────────┼──────────┤
 │ (0.0,0.0,20) 1024 3-slice │ 20.1µs │ 23.3µs   │
 ├───────────────────────────┼────────┼──────────┤
 │ (0.0,0.0,20) 1024 4-slice │ 21.7µs │ 24.0µs   │
 └───────────────────────────┴────────┴──────────┘

With a cherry pick of the fix in #9562, we get

 ┌────────────────────────────┬───────┬──────┬────────┬───────────────┬─────────────┬───────────┐
 │ Benchmark                  │ Nulls │ Rows │ Arrays │ Fallback (µs) │ Native (µs) │ Change    │
 ├────────────────────────────┼───────┼──────┼────────┼───────────────┼─────────────┼───────────┤
 │ list_view<i64>(0.1,0.1,20) │ 10%   │ 100  │ 3      │ 4.19          │ 2.93        │ -30.1%    │
 ├────────────────────────────┼───────┼──────┼────────┼───────────────┼─────────────┼───────────┤
 │ list_view<i64>(0.1,0.1,20) │ 10%   │ 400  │ 3      │ 10.85         │ 10.04       │ -9.4%     │
 ├────────────────────────────┼───────┼──────┼────────┼───────────────┼─────────────┼───────────┤
 │ list_view<i64>(0.1,0.1,20) │ 10%   │ 1024 │ 3      │ 27.80         │ 26.67       │ -5.2%     │
 ├────────────────────────────┼───────┼──────┼────────┼───────────────┼─────────────┼───────────┤
 │ list_view<i64>(0.1,0.1,20) │ 10%   │ 1024 │ 4      │ 28.06         │ 25.06       │ -9.7%     │
 ├────────────────────────────┼───────┼──────┼────────┼───────────────┼─────────────┼───────────┤
 │ list_view<i64>(0.0,0.0,20) │ 0%    │ 100  │ 3      │ 2.57          │ 1.42        │ -45.0%    │
 ├────────────────────────────┼───────┼──────┼────────┼───────────────┼─────────────┼───────────┤
 │ list_view<i64>(0.0,0.0,20) │ 0%    │ 400  │ 3      │ 5.78          │ 7.17        │ +22.3%    │
 ├────────────────────────────┼───────┼──────┼────────┼───────────────┼─────────────┼───────────┤
 │ list_view<i64>(0.0,0.0,20) │ 0%    │ 1024 │ 3      │ 13.34         │ 18.96       │ +22.2%    │
 ├────────────────────────────┼───────┼──────┼────────┼───────────────┼─────────────┼───────────┤
 │ list_view<i64>(0.0,0.0,20) │ 0%    │ 1024 │ 4      │ 13.85         │ 17.93       │ +13.5%    │
 └────────────────────────────┴───────┴──────┴────────┴───────────────┴─────────────┴───────────┘

[brancz]

Do you mind comparing this to the fallthrough performance of #9562 ?

[vegarsti]

Do you mind comparing this to the fallthrough performance of #9562 ?

Oh for sure, thanks for reminding me!

[vegarsti]

Updated the description with results now. It's not looking like a win..!

[brancz]

I would say let's merge the fallthrough and iterate on this version. I'm sure there are several possibilities for optimizations.

[asubiotto]

FWIW I pushed up the branch I've had marinating locally for a month or two in case it's helpful: main...polarsignals:arrow-rs:asubiotto/lvinterleave. I believe the benchmarks showed a slight regression for interleaves of small lists, but overall the perf was an improvement. I'm not able to take a closer look right now, but sharing in case it's helpful.