Add scalar function performance baselines

[connortsui20]

Tracking Issue: #9128

Adds stable scalar function performance baselines for byte_length, l2_norm, l2_denorm, inner_product, cosine_similarity, GeoContains, and GeoIntersects.

The benchmarks use the public construction and execution paths, with no RowFn internals or forced execution strategies. The tensor arms vary width and nullability. The cosine arms also cover both constant-vector representations. The geo arms cover column and constant operands, mostly-disjoint and mostly-overlapping polygons, and nullable inputs.

The l2_denorm arms measure decoding the L2Denorm array encoding rather than a scalar function, since #9138 reclassified it. The benchmark and arm names are unchanged so CodSpeed can still line them up against develop.

The benchmark sizes are all comfortably above the roughly 146-instruction Divan harness floor from #9011. Each benchmark binary also uses vendored mimalloc, since scalar function execution allocates its output inside the timed trace and glibc differences between runner images caused the flakes fixed in #8861. vortex-tensor and vortex-geo run in a dedicated CodSpeed shard.

CodSpeed can only compare a later implementation change when the same benchmark name already exists on develop, so these baselines need to land before the RowFn ports. This PR is stacked on the constructor standardization because the ordinary scalar functions use ScalarFnFactoryExt.

Checks:

• cargo check -p vortex-array -p vortex-tensor -p vortex-geo --benches
• cargo bench -p vortex-array --bench byte_length -- --test
• cargo bench -p vortex-tensor --bench cosine_similarity --bench l2_norm --bench l2_denorm --bench inner_product -- --test
• cargo bench -p vortex-geo --bench binary_predicates -- --test
• two repeated sample runs for byte_length and all tensor benchmarks (byte-length medians stayed within 0.4%, tensor medians stayed within 3.3%)
• cargo clippy -p vortex-array -p vortex-tensor -p vortex-geo --all-targets --all-features
• cargo clippy --all-targets --all-features
• cargo +nightly fmt --all
• yamllint --strict -c .yamllint.yaml .github/workflows/codspeed.yml
• 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.

✅ 1885 untouched benchmarks
🆕 53 new benchmarks
⏩ 12 skipped benchmarks¹

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
🆕	Simulation	inline[4096]	N/A	61.6 µs	N/A
🆕	Simulation	inline[65536]	N/A	787.6 µs	N/A
🆕	Simulation	nullable_out_of_line[4096]	N/A	73.8 µs	N/A
🆕	Simulation	nullable_out_of_line[65536]	N/A	793.8 µs	N/A
🆕	Simulation	out_of_line[4096]	N/A	61.8 µs	N/A
🆕	Simulation	out_of_line[65536]	N/A	787.6 µs	N/A
🆕	Simulation	column_x_column_points	N/A	14.1 ms	N/A
🆕	Simulation	column_x_column_polygons	N/A	23.6 ms	N/A
🆕	Simulation	constant_x_nullable_points	N/A	12.6 ms	N/A
🆕	Simulation	constant_x_nullable_polygons_disjoint	N/A	17.9 ms	N/A
🆕	Simulation	constant_x_points	N/A	13.7 ms	N/A
🆕	Simulation	constant_x_polygons_disjoint	N/A	17.8 ms	N/A
🆕	Simulation	constant_x_polygons_overlapping	N/A	2 s	N/A
🆕	Simulation	column_x_column_polygons	N/A	22.1 ms	N/A
🆕	Simulation	nullable_points_x_constant	N/A	12.7 ms	N/A
🆕	Simulation	nullable_polygons_disjoint_x_constant	N/A	16 ms	N/A
🆕	Simulation	points_x_constant	N/A	13.7 ms	N/A
🆕	Simulation	polygons_disjoint_x_constant	N/A	15.6 ms	N/A
🆕	Simulation	polygons_overlapping_x_constant	N/A	26.8 ms	N/A
🆕	Simulation	multipoint_non_nullable	N/A	65.5 ms	N/A
...	...	...	...	...	...

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

---

_{Comparing ct/scalar-fn-baselines (880ddac) with ct/scalar-fn-factory-ext (bf43ecf)}

Footnotes

1. 12 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. ↩

[joseph-isaacs]

This benchmark has a too long runtime