chore: deny clippy::inline_always at the workspace level

[joseph-isaacs]

Summary

Ban new uses of #[inline(always)] across the workspace. clippy::inline_always is pedantic-only and the workspace does not enable the pedantic group, so the lint has to be named explicitly to have any effect.

The 91 existing uses are deliberate hot-path annotations — vortex-buffer bit accessors, the SIMD compress and AVX2 take kernels, vortex-compute lane kernels, FastLanes unpacking — several with comments explaining why always is load-bearing (e.g. #[target_feature] wrappers that need a fully-inlined copy per feature set). Removing them would change codegen in the hottest paths in the repo with no benchmarks to back it, so they are grandfathered with a per-site #[allow(clippy::inline_always)] instead. Codegen is unchanged. Any new #[inline(always)] now fails the build unless the author adds the same explicit opt-out, which is the point.

Changes

• Cargo.toml: inline_always = "deny" in [workspace.lints.clippy], with a comment explaining the grandfathering.
• 36 files: added #[allow(clippy::inline_always)] above each existing #[inline(always)]. Mechanical, no behavior or codegen change.

Checks

• cargo clippy --workspace --all-targets --all-features — clean.
  • vortex-duckdb, vortex-sqllogictest and benchmarks/duckdb-bench were excluded locally: the vortex-duckdb build script downloads duckdb v1.5.5.zip and gets HTTP 403 through the sandbox proxy. None of those three crates contains an #[inline(always)], so they need no grandfathering; CI covers them.
• cargo +nightly fmt --all — clean (unrelated reformatting this newer nightly wanted in vortex-ffi / vortex-duckdb was reverted, so it stays out of this diff).
• Verified the ban bites: a temporary #[inline(always)] added to vortex-utils failed cargo clippy -p vortex-utils, then was removed.
• No tests run — the change adds only lint attributes.

🤖 Generated with Claude Code

https://claude.ai/code/session_01GTBhQJigN9KMk4nAnRFCzs

[codspeed-hq]

Merging this PR will degrade performance by 26.24%

⚠️ 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
✅ 2092 untouched benchmarks
⏩ 206 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	WallTime	arrow_checked_add_u32_neon[16384]	13.4 µs	20.4 µs	-34.36%
❌	WallTime	arrow_checked_add_u32_avx2[16384]	17.7 µs	21.3 µs	-17.1%

Tip

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

---

_{Comparing claude/vortex-inline-always-lint-iuc5no (e26d5b9) with develop (b4863a1)}

Footnotes

1. 206 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]

I would expect these inlines are not always necessary but let's grandfather them for now