Add fixed-size binary support

[connortsui20]

Stacked on #8864.

Why

FixedSizeBinary has the same one-buffer fixed-width structure, so it should inherit the shared structural compute rather than introduce a third take/filter implementation. It remains a distinct logical type and stores an explicit row count because zero-width values cannot derive length from the values buffer.

What

• add dtype, scalar, canonical array, builder, serialization, Arrow, Python, FFI, CUDA, file, fuzz, and conformance support
• implement the same three-hook FixedWidthArray adapter used by Primitive and Decimal; FSB has no local take or filter pipeline
• reuse byte-compress filtering through 4 bytes, typed filtering through 32 bytes, and the common runtime-width fallback
• reuse AVX2 takes for 4- and 8-byte records and common scalar take paths for other widths
• validate physical buffer lengths and preserve zero-width row counts through Vortex and Arrow

Issue provenance

• Existing and fixed: compact Serde decoding had stale numeric positions for Extension and Union.
• Introduced during development and fixed: compact-tag placement, missing ArrayParts validation, zero-width Arrow handling, and full-source cloning for uncommon-width filters.

Validation

• cargo test -p vortex-array --lib (3,064 passed, 1 ignored)
• cargo clippy -p vortex-array --all-targets --all-features
• cargo test -p vortex-arrow fixed_size_binary
• cargo test -p vortex-file test_round_trip_many_types
• cargo test --no-run --target x86_64-apple-darwin -p vortex-array --lib
• 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.

✅ 1884 untouched benchmarks
⏩ 12 skipped benchmarks¹

---

_{Comparing ct/fixed-width-encoding (3935ac9) with ct/fixed-width-primitive-decimal (dfc6a40)}

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. ↩

[github-actions]

Polar Signals Profiling Results

Latest Run

Status	Commit	Job	Attempt	Link
🟢 Done	460d6dc		1	Explore Profiling Data

Previous Runs (9)

Status	Commit	Job	Attempt	Link
🟢 Done	935e86c		1	Explore Profiling Data
🟢 Done	11a4e4f		1	Explore Profiling Data
🟢 Done	d36c772		1	Explore Profiling Data
🟢 Done	c924a06		1	Explore Profiling Data
🟢 Done	97c8e02		1	Explore Profiling Data
🟢 Done	0592a4f		1	Explore Profiling Data
🟢 Done	8dedc68		1	Explore Profiling Data
🟢 Done	8ef7bb2		1	Explore Profiling Data
🟢 Done	5dd799a		1	Explore Profiling Data

---

Powered by Polar Signals Cloud

[github-actions]

Benchmarks: Vortex queries 📖

Verdict: No clear signal (low confidence)
Attributed Vortex impact: +2.0%
Engines: DataFusion No clear signal (+6.1%, low confidence) · DuckDB No clear signal (-1.9%, low confidence)
Vortex (geomean): 1.026x ➖
Parquet (geomean): 1.016x ➖
Shifts: Parquet (control) +1.6% · Median polish +1.6%

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 (1.085x ➖, 0↑ 1↓)

name	PR 460d6dc (ns)	base 57962f4 (ns)	ratio (PR/base)
vortex_q00/datafusion:vortex-file-compressed	8648179	8197567	1.05
vortex_q01/datafusion:vortex-file-compressed 🚨	5452852	4885078	1.12

datafusion / parquet (1.023x ➖, 0↑ 0↓)

name	PR 460d6dc (ns)	base 57962f4 (ns)	ratio (PR/base)
vortex_q00/datafusion:parquet	20848141	20638806	1.01
vortex_q01/datafusion:parquet	4744100	4580085	1.04

duckdb / vortex-file-compressed (0.989x ➖, 0↑ 0↓)

name	PR 460d6dc (ns)	base 57962f4 (ns)	ratio (PR/base)
vortex_q00/duckdb:vortex-file-compressed	10192401	10061599	1.01
vortex_q01/duckdb:vortex-file-compressed	5960426	6167184	0.97

duckdb / parquet (1.009x ➖, 0↑ 0↓)

name	PR 460d6dc (ns)	base 57962f4 (ns)	ratio (PR/base)
vortex_q00/duckdb:parquet	23424626	23195009	1.01
vortex_q01/duckdb:parquet	9413805	9340266	1.01

---

File Size Changes (1 files changed, +0.0% overall, 1↑ 0↓)

File	Scale	Format	Base	HEAD	Change	%
test.vortex	1.0	vortex-file-compressed	80.59 MB	80.59 MB	+48 B	+0.0%

Totals:

• vortex-file-compressed: 80.85 MB → 80.85 MB (+0.0%)

[codecov]

Codecov Report

❌ Patch coverage is 64.16235% with 415 lines in your changes missing coverage. Please review.
✅ Project coverage is 82.17%. Comparing base (fa4beb4) to head (cc89366).

Files with missing lines	Patch %	Lines
...rtex-array/src/arrays/fixed_size_binary/compute.rs	30.28%	99 Missing ⚠️
vortex-array/src/arrays/fixed_size_binary/mod.rs	77.05%	53 Missing ⚠️
vortex-array/src/builders/fixed_size_binary.rs	54.16%	44 Missing ⚠️
encodings/sparse/src/canonical.rs	0.00%	41 Missing ⚠️
...rtex-array/src/arrays/constant/vtable/canonical.rs	0.00%	17 Missing ⚠️
...x-array/src/scalar_fn/fns/binary/compare/nested.rs	0.00%	17 Missing ⚠️
vortex-array/src/arrays/dict/execute.rs	0.00%	14 Missing ⚠️
vortex-datafusion/src/convert/scalars.rs	33.33%	14 Missing ⚠️
vortex-array/src/dtype/serde/serde.rs	53.57%	13 Missing ⚠️
.../aggregate_fn/fns/is_constant/fixed_size_binary.rs	0.00%	12 Missing ⚠️
... and 26 more

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[joseph-isaacs]

please no new DType