GPU: materialize flat constant arrays on device

[joseph-isaacs]

Independent of the other follow-up PRs; based directly on #9147. This draft contains one reviewable GPU correctness or performance concern.

What

• Materialize null, boolean, primitive, decimal, UTF-8, binary, and flat extension constants as device-resident canonical arrays.
• Preserve nullable validity, including all-invalid constants.
• Add focused coverage for inline and outlined values.

Why this is needed

Constant nodes are normal compressor output: constant columns, all-null columns, and constant patch values all produce ConstantArray. The base CUDA executor only handles non-null numeric/decimal constants, so Boolean, UTF-8, binary, extension, and null constants fall back to CPU inside an otherwise GPU-resident tree. The profile exposed a concrete “Nullable UTF-8 Constant CPU fallback” costing 14.573 ms, making this both a correctness-coverage and measured performance fix.

Validation

Checked independently against this PR's current base:

• cargo check -p vortex-cuda --all-features

The combined implementation also passed Rust/CUDA formatting, 164 focused dynamic-dispatch tests, and 28 focused constant-array tests.

[codspeed-hq]

Merging this PR will degrade performance by 11.78%

⚠️ 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.

❌ 2 regressed benchmarks
✅ 2041 untouched benchmarks
⏩ 46 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	WallTime	words_gather_scalar[65536]	8.2 µs	9.4 µs	-12.45%
❌	WallTime	words_gather_dispatch[1024]	8 ns	9 ns	-11.11%

Tip

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

---

_{Comparing joe/gpu-flat-constant-arrays (ee5e383) with develop (b825c4f)}

Footnotes

1. 46 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. ↩