refactor(array): pass the execution context into patch lookups

[joseph-isaacs]

Summary

Stacked on #9904. Follows up the review there by threading the caller's execution context through every patch lookup, rather than only into search_index_chunked.

Patches::get_patched and search_index each built their own execution context from the legacy session, even though every caller of get_patched is a scalar_at that already holds one.

Changes

• Patches::get_patched, search_index, search_index_chunked and the private binary-search helpers take ctx: &mut ExecutionCtx instead of calling legacy_session().create_execution_ctx(). The clippy::disallowed_methods allowances on those paths go away.
• The four scalar_at callers (ALP, ALP-RD, BitPacked, Sparse) pass their context through; two of them previously bound it as _ctx.
• Patches::slice has no caller context, so it creates one legacy context at the top and reuses it for both index searches and the chunk base read.
• The patches_lookup bench and the patches tests create a context once and pass it in.

API Changes

Patches::get_patched and Patches::search_index gain a ctx: &mut ExecutionCtx parameter.

🤖 Generated with Claude Code

https://claude.ai/code/session_01ACe1mztjw9vNFUYwzVhGwz

---

Generated by Claude Code

[codspeed]

Merging this PR will regress 7 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.

⚠️ 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

⚡ 52 improved benchmarks
❌ 7 regressed benchmarks
✅ 112 untouched benchmarks
⏩ 2366 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
❌	filtered_sink_i64_avx512[OneNullInEight]	22.3 µs	31.6 µs	-29.42%
❌	words_gather_dispatch_avx512[65536]	987 ns	1,343 ns	-26.51%
❌	filtered_sink_i64_avx2[NineNullsInTen]	14.8 µs	18.2 µs	-18.8%
❌	filtered_sink_i64_avx512[NineNullsInTen]	15 µs	18.3 µs	-17.67%
❌	filtered_sink_i64_avx2[OneNullInEight]	26.1 µs	31.1 µs	-15.99%
❌	multiply_shapes_neon[(16384, PerRowPerRow)]	17.4 µs	20.3 µs	-14.09%
❌	words_gather_scalar_avx2[65536]	8.2 µs	9.4 µs	-11.98%
⚡	mul_i16_nonnull_avx2	28.9 µs	11.6 µs	×2.5
⚡	mul_i32_nonnull_avx2	32.9 µs	13.4 µs	×2.5
⚡	mul_i8_nonnull_avx2	29.9 µs	12.6 µs	×2.4
⚡	mul_u8_nonnull_avx512	22.5 µs	9.6 µs	×2.3
⚡	mul_i32_nullable_avx2	34.4 µs	14.9 µs	×2.3
⚡	mul_i32_nonnull_neon	23.8 µs	10.4 µs	×2.3
⚡	mul_i8_nonnull_neon	25.7 µs	11.2 µs	×2.3
⚡	mul_i16_nonnull_neon	24.2 µs	10.6 µs	×2.3
⚡	mul_u8_nonnull_neon	19 µs	8.4 µs	×2.3
⚡	mul_u32_nonnull_neon	18.4 µs	8.1 µs	×2.3
⚡	mul_u64_nonnull_avx2	39.3 µs	17.4 µs	×2.3
⚡	mul_u16_nonnull_neon	18.4 µs	8.2 µs	×2.2
⚡	add_i32_nonnull_neon	17.4 µs	7.9 µs	×2.2
...	...	...	...	...

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

Tip

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

---

_{Comparing claude/context-passing-parent-fo4x8u (c550195) with ji/patches-chunk-offset-probe (2e0df78)}

Footnotes

1. 2366 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. ↩