Experiment: compose scheme registration with output editions

[connortsui20]

Summary

• Tracking Issue: Tracking Issue: Scheme Design with Editions #10038
• Experimental alternative to Register compression schemes on the session #10034, built from its current head.

Makes scheme registration the writer's opt-in to compression. Encoding packages own their schemes, and compressor construction resolves those registrations against the permitted serialized IDs before starting the cascading engine. Registering an encoding alone still supports reading without enabling its compression scheme.

Changes

• Move encoding-specific schemes into their owning packages. Keep vortex-btrblocks as a thin initialization and construction facade, with compatibility re-exports.
• Replace the default scheme list and scheme builder with explicit, atomic session registration. Compact is an additive initialization helper. A compression-session fork isolates per-operation registrations.
• Declare direct helper outputs as well as top-level encodings. Constants become registered schemes, and the default writer passes its exact serialization permissions to construction.
• Add a complete draft CUDA edition and explicit edition-set replacement. CUDA writers select it instead of core. Enabling several editions still unions their permissions.

Scoped Clippy passed with warnings denied for the affected core and encoding crates. Compiler checks passed for the affected core crates, including tests, benches, examples, all-feature compression builds, and the facade without file support. Added regression coverage for registration, permissions, concurrency, and edition selection. Rust and TOML formatting passed. Tests were not executed. Snapshot review, CUDA execution, Wasm execution, and benchmarks remain unrun. The whole-workspace pre-push Clippy hook was skipped after scoped checks. Stateful training in #8434 remains separate.

API Changes

Removes BtrBlocksCompressorBuilder, with_compact, and only_cuda_compatible. Call package initialization functions, select editions, then construct BtrBlocksCompressor::from_session. The layout builder remains and accepts an already constructed compressor through with_btrblocks_compressor.

VortexSession::clone continues to share registrations. fork_compression copies the compression registry for local configuration while sharing other services. register_cuda_layout only registers support, while cuda_write_strategy selects the CUDA edition on the supplied session. The cascading engine retains sampling, estimates, exclusions, and recursive compression.

[connortsui20]

this is bad

[codspeed]

Merging this PR will regress 2 benchmarks

⚠️ 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.

⚠️ 3 benchmarks measured no execution time

Nothing ran under measurement, usually because the compiler removed the code under test. These results are not comparable, so they count as unchanged.

Preventing compiler optimizations

⚠️ Different runtime environments detected

Some benchmarks with significant performance changes were compared across different runtime environments,
which may affect the accuracy of the results.

Open the report in CodSpeed to investigate

⚡ 2 improved benchmarks
❌ 2 regressed benchmarks
✅ 2218 untouched benchmarks
⏩ 343 skipped benchmarks¹
🗄️ 1 archived benchmark run²

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	WallTime	decode_avx512[8192, (Inline, OneNullInEight)]	81.7 µs	99.2 µs	-17.65%
❌	WallTime	decode_avx512[8192, (Inline, AllValid)]	96.6 µs	112.5 µs	-14.12%
⚡	Simulation	from_vec_drop_arrow[16384]	40.4 µs	31.4 µs	+28.7%
⚡	WallTime	filtered_sink_i64_avx512[OneNullInEight]	26.3 µs	22.3 µs	+18.26%
⚠️	Simulation	fixed_16_advancing_ptr_safe[100]	< 1 ns	< 1 ns	N/A
⚠️	Simulation	preverify_advancing_ptr_unchecked[1000]	< 1 ns	< 1 ns	N/A
⚠️	Simulation	preverify_advancing_ptr_unchecked[10000]	< 1 ns	< 1 ns	N/A

Tip

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

---

_{Comparing ct/composable-schemes (ffe45b2) with develop (035aeff)}

Footnotes

1. 343 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. ↩

2. 1 benchmark was run, but is now archived. If it was deleted in another branch, consider rebasing to remove it from the report. Instead if it was added back, click here to restore it. ↩