feat(array): builders no longer canonicalize their children [builders-child-stack]

[robert3005]

Builders are lazy. Fully canonical builders force a lot of unnecessary decompression which goes against what vortex aims for. With this change builders behave more like canonical arrays where the top level isn't lazy but the children are. Practically speaking this means that all nested type builders store lazy children and validity

[codspeed-hq]

Merging this PR will improve performance by 46.53%

⚡ 8 improved benchmarks
✅ 1921 untouched benchmarks
⏩ 51 skipped benchmarks¹

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
⚡	Simulation	extend_from_array_zctl[(1000, 64)]	2,184.4 µs	722.9 µs	×3
⚡	Simulation	extend_from_array_zctl[(10000, 8)]	3.6 ms	1.8 ms	×2
⚡	Simulation	decompress[u64, (1000, 16)]	130 µs	73.1 µs	+77.78%
⚡	Simulation	extend_from_array_zctl[(1000, 8)]	903.1 µs	715.5 µs	+26.21%
⚡	Simulation	chunked_dict_primitive_canonical_into[f32, (1000, 100, 100)]	1.7 ms	1.5 ms	+13.09%
⚡	Simulation	extend_from_array_non_zctl_overlapping[(10000, 8)]	2.1 ms	1.8 ms	+12.77%
⚡	Simulation	chunked_dict_primitive_canonical_into[u32, (1000, 10, 100)]	1.6 ms	1.5 ms	+11.11%
⚡	Simulation	search_index_in_range_chunked	5.9 ms	5.3 ms	+10.12%

Tip

Curious why this is faster? Comment @codspeedbot explain why this is faster on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing claude/builders-canonical-children-9ze0t6 (76b988c) with claude/builders-bulk-loop-callers-9ze0t6 (e570927)}

Footnotes

1. 51 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. ↩

[robert3005]

This is the meat of this change

[robert3005]

Collapsed into #9064 along with #8966 and #9131.