Fix flat reader subrange decode reuse

[lukekim]

Fixes #8587.

Summary

• Memoize FlatReader's decoded array future so synthetic subrange scans share the decoded flat segment instead of issuing repeated decode work.
• Add regression coverage for the query patterns that regressed after perf(scan): intra-file decode parallelism — sub-split large chunk spans #8400, including projection-only, filter-only, filtered projection, computed projection, string filtered projection, and string filtered computed projection cases.

Validation

• cargo nextest run -p vortex-layout -E 'test(layouts::flat::reader)'
• cargo nextest run -p vortex-layout
• cargo clippy -p vortex-layout --all-targets --all-features
• git diff --check
• cargo bench --workspace
• Re-ran SQL benches with /opt/homebrew/bin/uv 0.11.24; core suites passed: Appian, TPCH, TPCDS, ClickBench, ClickBench sorted, FineWeb, and GH Archive via direct binary rerun. PolarSignals/StatPopGen exposed pre-existing benchmark-definition/runtime backend failures, and bare Public BI requires --opt dataset=<name>.

[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.

⚡ 5 improved benchmarks
❌ 3 regressed benchmarks
✅ 1581 untouched benchmarks
⏩ 4 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Simulation	chunked_bool_canonical_into[(1000, 10)]	15.9 µs	26.7 µs	-40.29%
❌	Simulation	chunked_varbinview_into_canonical[(1000, 10)]	169.1 µs	205.8 µs	-17.83%
❌	Simulation	slice_empty_vortex	310 ns	368.3 ns	-15.84%
⚡	Simulation	bitwise_not_vortex_buffer_mut[128]	273.6 ns	215.3 ns	+27.1%
⚡	Simulation	bitwise_not_vortex_buffer_mut[1024]	333.9 ns	275.6 ns	+21.17%
⚡	Simulation	bitwise_not_vortex_buffer_mut[2048]	427.8 ns	369.4 ns	+15.79%
⚡	Simulation	chunked_varbinview_canonical_into[(100, 100)]	259.6 µs	224.5 µs	+15.65%
⚡	Simulation	chunked_varbinview_into_canonical[(100, 100)]	306.8 µs	271.9 µs	+12.84%

Tip

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

---

_{Comparing spiceai:lukim/8587-regression (0b86845) with develop (bdbf6c4)}

Footnotes

1. 4 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]

We had this at one point and removed it #4459 because this effectively doubles memory usage of the reader. Any kind of caching like this has to have invalidation logic. We need memory benchmarks to flag changes like these but we have investigated this in the past