feat(row): support remaining ordered dtypes

[gatesn]

Summary

• add Decimal256 row keys using the signed 256-bit order-preserving transform
• add prefix-safe lexicographic encoding for variable-size lists and maps
• keep Extension rejected until extension dtypes can explicitly declare storage-order equivalence
• add reproducible List<i64> and List<Utf8> throughput and output-size benchmarks against arrow-row
• update the row fuzz oracle and byte-format specification

Stack

• feat(row): support remaining ordered dtypes #9553 (this PR): dtype support and benchmark baseline
• perf(row): compact variable-list keys #9554: compact variable-list keys and reuse sizing-pass preparation

Baseline benchmarks

cargo bench -p vortex-row --bench row_encode -- --output-sizes

Case	arrow-row bytes	Vortex bytes
List<i64>	153,000	146,000
List<Utf8>	256,347	546,000

Local median encode times with 1,000 lists of 8 elements:

Case	arrow-row	Vortex
List<i64>	36.83 us	46.74 us
List<Utf8>	94.74 us	100.7 us

Testing

• cargo test -p vortex-row
• cargo check -p vortex-row --bench row_encode
• cargo clippy -p vortex-row --all-targets --all-features
• cargo check -p vortex-fuzz --no-default-features --lib
• cargo +nightly fmt --all -- --check
• git diff --check

[codspeed-hq]

Merging this PR will not alter performance

⚠️ 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.

✅ 1981 untouched benchmarks
🆕 4 new benchmarks
⏩ 54 skipped benchmarks¹

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
🆕	Simulation	list_i64_arrow_row	N/A	1.7 ms	N/A
🆕	Simulation	list_i64_vortex	N/A	2.2 ms	N/A
🆕	Simulation	list_utf8_arrow_row	N/A	3.4 ms	N/A
🆕	Simulation	list_utf8_vortex	N/A	4.1 ms	N/A

---

_{Comparing ngates/vortex-row-dtype-support (41da939) with develop (332ae0b)}

Footnotes

1. 54 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. ↩