Add PiecewiseSequence take microbenchmarks

[danking]

@robert3005 I'm not sure if you want to check these in. These are what I used to reason about various ways to copy slices of data from one spot in memory to another.

Summary

• add piecewise_sequence_take_primitive to compare constant-length PiecewiseSequence take against dense lengths and reifying constant lengths
• add take_slices_to_buffer_matrix to compare extend/cursor-copy loops with and without preverification plus unchecked slicing

Validation

• cargo +nightly fmt --all
• cargo clippy -p vortex-array --all-targets --all-features -- -D warnings
• cargo bench -p vortex-array --bench piecewise_sequence_take_primitive
• cargo bench -p vortex-array --bench take_slices_to_buffer_matrix

[robert3005]

I think we want benchmarks. I came up with the long term solution in #8825 and we have to make sure we don't regress

[codspeed-hq]

Merging this PR will not alter performance

✅ 1788 untouched benchmarks
🆕 39 new benchmarks
⏩ 54 skipped benchmarks¹

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
🆕	Simulation	optimized_constant_lengths[1]	N/A	427.3 µs	N/A
🆕	Simulation	optimized_constant_lengths[16]	N/A	102.3 µs	N/A
🆕	Simulation	optimized_constant_lengths[256]	N/A	87.6 µs	N/A
🆕	Simulation	optimized_constant_lengths[4]	N/A	149.5 µs	N/A
🆕	Simulation	optimized_constant_lengths[64]	N/A	89.9 µs	N/A
🆕	Simulation	prebuilt_dense_lengths[1]	N/A	456.9 µs	N/A
🆕	Simulation	prebuilt_dense_lengths[16]	N/A	103.7 µs	N/A
🆕	Simulation	prebuilt_dense_lengths[256]	N/A	87.5 µs	N/A
🆕	Simulation	prebuilt_dense_lengths[4]	N/A	156.5 µs	N/A
🆕	Simulation	prebuilt_dense_lengths[64]	N/A	90 µs	N/A
🆕	Simulation	reify_constant_lengths_then_take[1]	N/A	471.6 µs	N/A
🆕	Simulation	reify_constant_lengths_then_take[16]	N/A	110.2 µs	N/A
🆕	Simulation	reify_constant_lengths_then_take[256]	N/A	93.5 µs	N/A
🆕	Simulation	reify_constant_lengths_then_take[4]	N/A	163.3 µs	N/A
🆕	Simulation	reify_constant_lengths_then_take[64]	N/A	96.4 µs	N/A
🆕	Simulation	cursor_copy_safe[1000]	N/A	6.9 µs	N/A
🆕	Simulation	cursor_copy_safe[10000]	N/A	14.2 µs	N/A
🆕	Simulation	cursor_copy_safe[100000]	N/A	86.6 µs	N/A
🆕	Simulation	extend_safe[1000]	N/A	7.4 µs	N/A
🆕	Simulation	extend_safe[10000]	N/A	14 µs	N/A
...	...	...	...	...	...

ℹ️ Only the first 20 benchmarks are displayed. Go to the app to view all benchmarks.

---

_{Comparing codex/piecewise-sequence-benchmarks (c731e4f) with develop (ae9ecec)}

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. ↩

[robert3005]

I think in order to merge them we have to make them smaller, these should really be fast microbenchmarks

[github-actions]

Polar Signals Profiling Results

Latest Run

Status	Commit	Job	Attempt	Link
🟢 Done	c731e4f		1	Explore Profiling Data

Previous Runs (1)

Status	Commit	Job	Attempt	Link
🟢 Done	9995892		1	Explore Profiling Data

---

Powered by Polar Signals Cloud

[github-actions]

Benchmarks: Vortex queries 📖

Verdict: No clear signal (medium confidence)
Attributed Vortex impact: +8.8%
Engines: DataFusion No clear signal (+7.2%, environment too noisy confidence) · DuckDB Likely regression (+10.4%, medium confidence)
Vortex (geomean): 1.174x ❌
Parquet (geomean): 1.125x ❌
Shifts: Parquet (control) +12.5% · Median polish +18.3%

How to read Verdict and Engines

• Verdict: Overall PR-level signal after subtracting baseline drift estimated from Parquet control rows. It can be Likely improvement, Likely regression, or No clear signal.
• Engines: Per-engine attribution. DataFusion is compared against DataFusion/Parquet controls; DuckDB is compared against DuckDB/Parquet controls. This answers whether each engine improved or regressed independently.
• Confidence: Based on directional consistency, share of rows above the noise floor, and control-run noise.

---

datafusion / vortex-file-compressed (1.272x ❌, 0↑ 2↓)

name	PR c731e4f (ns)	base ae9ecec (ns)	ratio (PR/base)
vortex_q00/datafusion:vortex-file-compressed 🚨	11716525	9244952	1.27
vortex_q01/datafusion:vortex-file-compressed 🚨	7752580	6075932	1.28

datafusion / parquet (1.186x ❌, 0↑ 2↓)

name	PR c731e4f (ns)	base ae9ecec (ns)	ratio (PR/base)
vortex_q00/datafusion:parquet 🚨	23121989	20820621	1.11
vortex_q01/datafusion:parquet 🚨	5925270	4679197	1.27

duckdb / vortex-file-compressed (1.179x ❌, 0↑ 2↓)

name	PR c731e4f (ns)	base ae9ecec (ns)	ratio (PR/base)
vortex_q00/duckdb:vortex-file-compressed 🚨	11781191	10294945	1.14
vortex_q01/duckdb:vortex-file-compressed 🚨	7503734	6173707	1.22

duckdb / parquet (1.068x ➖, 0↑ 0↓)

name	PR c731e4f (ns)	base ae9ecec (ns)	ratio (PR/base)
vortex_q00/duckdb:parquet	25560487	23450715	1.09
vortex_q01/duckdb:parquet	9926883	9489168	1.05

---

No file size changes detected.

[danking]

@robert3005 each one is <=500us now but I could bring them down further if you like.

[robert3005]

we generally held the bar at 1ms so this is good enough