Select Decimal compression format from writer permissions

[mhk197]

Summary

Enable compression of wide decimals when the writer permits vortex.decimal_byte_parts.v2. Writers restricted to v1 continue to leave wide decimals canonical.

Changes

• Keep Decimal v1 in all_schemes() and store static scheme references plus an optional allowlist on the builder.
• Store permissions with with_allowed_encodings. Repeated calls intersect; during build(), select the latest permitted Decimal version and filter unsupported schemes, including schemes registered after permissions were supplied. Without an allowlist, registered configurations are preserved.
• Compress the most significant and lower parts of wide decimals independently. Single-part decimals continue to serialize as v1.
• Apply the CUDA preset after supplying permissions so it removes Decimal v2 from the stored allowlist. Later intersections cannot re-enable v2.
• Migrate the writer, bindings, CLI, CUDA integration, and benchmarks to the deferred configuration API.
• Cover format selection, explicit registrations, intersecting restrictions, CUDA compatibility, child compression, and edition-aware file round trips.

API Changes

ALL_SCHEMES is replaced by all_schemes(). DecimalScheme::new(v2) replaces construction as a unit struct. retain_allowed_encodings is replaced by with_allowed_encodings, which applies permissions at build time and can change explicitly registered Decimal versions. Static scheme registration is preserved.

Supply the initial allowlist before applying the CUDA preset:

BtrBlocksCompressorBuilder::default()
    .with_allowed_encodings(&allowed_serialized_ids)
    .only_cuda_compatible()
    .build()

Validation

Reviewed the source diff. Tests, builds, linting, and formatting were not run, per repository instructions.

[codspeed]

Merging this PR will degrade performance by 17.32%

⚠️ 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
✅ 196 untouched benchmarks
⏩ 2303 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
❌	scalar_subtract_neon	12 µs	15.1 µs	-20.66%
❌	filtered_sink_i64_avx512[OneNullInEight]	22.8 µs	26.5 µs	-13.83%

Tip

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

---

_{Comparing mk/configurable-schemes (195c7ae) with develop (27a2f10)}

Footnotes

1. 2303 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. ↩