Record canonical Linux benchmark baseline

[tomsanbear]

Summary

• replace the pre-squash Linux comparison with an authoritative five-rotation run from canonical merge commit 2dc82e0
• regenerate the Linux-only report and graphics without pooling or fabricating macOS data
• record full provenance, quality gates, mixed CPU outcomes, and the dedicated-hardware macOS deferral in TODO.md

Validation

• five CPU rotations pinned to Ryzen 7950X3D CPU 0; all 12 audits report 100% idle and no host issues
• explicit system allocator; inherited tcmalloc preload cleared and recorded
• 110 CPU rows and 60 allocation rows; zero post-drop bytes and valid reallocation decomposition
• deterministic report regeneration
• 7 Python reporting tests
• 114 Rust workspace tests
• 49 doctests

macOS is intentionally not included and remains deferred until dedicated controllable Apple Silicon hardware is available.

[codspeed-hq]

Merging this PR will improve performance by 14.02%

⚠️ 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

⚡ 3 improved benchmarks
✅ 107 untouched benchmarks

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
⚡	SmallVec4[4]	174.4 ns	145.3 ns	+20.08%
⚡	SmallVec8[8]	277.5 ns	248.3 ns	+11.74%
⚡	SmallVec4[8]	307.8 ns	278.6 ns	+10.47%

Tip

Curious why this is faster? Comment @codspeedbot explain why this is faster on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing bench/canonical-linux-baseline (dadc2ec) with main (2dc82e0)}