parquet bench + cuda changes to scan clickbench

[onursatici]

Summary

Closes: #000

Testing

[onursatici]

this is a standalone uv script to do the same scan we do on gpu-scan-bench above, but for parquet instead of vortex

[onursatici]

I needed this to solve a panic on datetimeparts

[onursatici]

I don't know why I was getting a uint here on datetimeparts, so hacked around this

[onursatici]

this is an irrelevant fix that I did first, but turned out it is datetimeparts

[onursatici]

skip validation on release builds because they force a host copy

[codspeed-hq]

Merging this PR will degrade performance by 10.2%

⚡ 3 improved benchmarks
❌ 3 regressed benchmarks
✅ 948 untouched benchmarks
⏩ 1466 skipped benchmarks¹

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
⚡	Simulation	chunked_opt_bool_canonical_into[(10, 1000)]	1.6 ms	1.4 ms	+16.12%
⚡	Simulation	bench_many_nulls[0.5]	379.5 µs	342.8 µs	+10.72%
⚡	Simulation	bench_many_nulls[0.9]	547.5 µs	481.6 µs	+13.7%
❌	Simulation	patched_take_200k_dispersed	5.1 ms	5.6 ms	-10.14%
❌	Simulation	map_each[BufferMut<i32>, 128]	770.6 ns	858.1 ns	-10.2%
❌	Simulation	true_count_vortex_buffer[128]	1 µs	1.2 µs	-10.06%

---

_{Comparing os/parquet-bench (4be081e) with os/gpu-scan-bench (f08430d)}

Footnotes

1. 1466 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. ↩