Thread ExecutionCtx into VTable::validate

[robert3005]

Stacked on #8609.

Adds a ctx: &mut ExecutionCtx parameter to VTable::validate and wires it through Array::try_from_parts / ArrayInner::try_new, so encoding validation that needs canonicalization can use a real execution context instead of the hidden global LEGACY_SESSION.

Changes

• VTable::validate and all impls (vortex-array + every encoding crate) gain the new final ctx/_ctx parameter. Most validations are pure metadata checks and ignore it (_ctx).
• RunEnd and FSST validate previously minted a LEGACY_SESSION ctx via a TODO(ctx); they now consume the threaded ctx (removes 2 LEGACY_SESSION uses).
• try_from_parts callers thread an in-scope ctx/session where one exists (RunEnd/ALPRD try_new, deserialize/compress paths use the available session).
• Constructors with no context yet (e.g. StructArray::try_new, DictArray::try_new, PrimitiveArray::from_buffer_handle, ALP/parquet-variant/fastlanes/zstd try_new) mint LEGACY_SESSION.create_execution_ctx() as intermediate scaffolding (+17 sites). These are candidates for follow-up ctx threading (e.g. the ALP/ALPRD PR) and will be covered by the forthcoming LEGACY_SESSION allow-list.

Verification

• cargo build -p vortex-array --all-targets ✅
• cargo test -p vortex-array ✅ (3078 unit + 69 doctests)
• cargo test for encodings (runend/fsst/alp/fastlanes) ✅
• cargo clippy --all-targets --all-features -- -D warnings on vortex-array + all encodings ✅
• cargo +nightly fmt --all --check ✅
• Downstream crates (vortex-layout/file/scan/datafusion/…) build; they neither implement VTable::validate nor call try_from_parts, so they are unaffected.

(Workspace-wide cargo runs locally exclude the CUDA/duckdb/tpchgen crates, which need toolchains/network not available in this environment.)

🤖 Generated with Claude Code

https://claude.ai/code/session_01Mkrgj5SuJpaKBXK9jydtSc

---

Generated by Claude Code

[codspeed-hq]

Merging this PR will degrade performance by 40.76%

⚠️ 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.

⚡ 1 improved benchmark
❌ 8 regressed benchmarks
✅ 1586 untouched benchmarks
⏩ 4 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Simulation	take_indices[(1000, 4)]	49.4 µs	185.2 µs	-73.35%
❌	Simulation	take[sorted_linear_len16384_run8_take512]	66 µs	205.2 µs	-67.84%
❌	Simulation	take_fsl_nullable_random[16, 100]	72.3 µs	209.1 µs	-65.43%
❌	Simulation	chunked_bool_canonical_into[(1000, 10)]	16.2 µs	26.4 µs	-38.4%
❌	Simulation	take_filter_primitive_large_random_mask_random_indices[(2500, 25000)]	283.9 µs	422.2 µs	-32.75%
❌	Simulation	chunked_varbinview_opt_canonical_into[(1000, 10)]	169.6 µs	206 µs	-17.67%
❌	Simulation	chunked_varbinview_opt_into_canonical[(1000, 10)]	183 µs	219.6 µs	-16.66%
❌	Simulation	compress[(10000, 4)]	199 µs	227 µs	-12.36%
⚡	Simulation	chunked_varbinview_into_canonical[(1000, 10)]	205.7 µs	168.9 µs	+21.8%

Tip

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

---

_{Comparing claude/nice-hypatia-g68zgt-validate-ctx (c4687a7) with claude/nice-hypatia-g68zgt (f8f6a90)}

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. ↩