fix: omit zone-map stats that a nullable struct's nulls would corrupt

[joseph-isaacs]

Struct fields are written unmasked: a field still holds a value in rows where the containing struct is null, and the struct validity is written as a separate child. Each field's zone map was computed over that unmasked stream, so null_count undercounted the logical column (it was always 0 for a non-nullable field under a nullable struct) and nan_count counted values the column never shows. Pruning on s.x IS NULL then discarded zones that held matching rows.

Add LayoutWriterContext::with_nullable_ancestor, set by StructStrategy on the context it hands to field writers of a nullable struct, and have ZonedStrategy keep only min/max style aggregates under it. Those remain sound as bounds because the unmasked field is a superset of the logical values. Count aggregates are omitted rather than recorded wrong until they can be computed over the masked values.

Claude-Session: https://claude.ai/code/session_01HRQvL72sLkXczGCjvFWaMG

Summary

• Tracking Issue: #000

Changes

[codspeed]

Merging this PR will regress 1 benchmark

⚠️ 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.

⚠️ 1 benchmark measured no execution time

Nothing ran under measurement, usually because the compiler removed the code under test. This result is not comparable, so it counts as unchanged.

Preventing compiler optimizations

• bench_compare_sliced_dict_primitive[(3333, 10000)]

⚠️ 6 benchmarks spent significant time in system calls

System calls cannot be consistently instrumented, so they are not included in the measure, which understates the real cost. Please switch to the Walltime instrument to accurately measure system calls.

Measurement and system calls

• decompress[u16, (4000, 256)]
• multipolygon_mixed_validity
• take_varbin
• filter_run_end[len=4096_run=16_density=0.1]
• random_i128[0.01]
• bench_compare_sliced_dict_primitive[(3333, 10000)]

⚡ 5 improved benchmarks
❌ 1 regressed benchmark
✅ 2201 untouched benchmarks

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	WallTime	bitpack_blocked_compress_avx512	4.9 µs	5.7 µs	-13.94%
⚡	WallTime	threshold_neon[8]	4 µs	3 µs	+33.84%
⚡	WallTime	filter_neon[64]	3.5 µs	2.9 µs	+19.76%
⚡	WallTime	threshold_neon[80]	5.8 µs	4.8 µs	+19.15%
⚡	WallTime	threshold_neon[64]	5.3 µs	4.5 µs	+18.52%
⚡	WallTime	filter_neon[80]	3.5 µs	3.2 µs	+11.25%

Tip

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

---

_{Comparing ji/magical-bell-b84xbc (d00d4e2) with develop (9d5a844)}