Correct Linux benchmark baseline and active roadmap

[tomsanbear]

Corrects two measurement-integrity issues discovered while investigating the apparent JackVec iteration regression.

• Routes every vector through one shared non-inlined iteration kernel, eliminating per-implementation loop-code placement artifacts
• Requires an explicit allocator policy and records inherited/effective injection variables
• Replaces the uncontrolled tcmalloc Linux report with a five-rotation system-allocator run
• Regenerates the Markdown report and deterministic SVGs from the corrected data
• Records the rejected 16-byte-header direction and the evidence behind it
• Rewrites the active TODO roadmap around the remaining credible CPU, memory, downstream, and release work

Corrected Linux result: JackVec has 8 confidence-qualified wins, 5 equivalents, 3 inconclusive results, and 6 losses versus Vec. Long sequential iteration is 1.000x Vec and equivalent across all five implementations.

Validation performed locally: reporting unit tests, comparison adapter tests, all-target comparison check, deterministic graphics regeneration, and whitespace checks. The physical run used Rust 1.97.0, CPU 0 of the Ryzen 7950X3D V-cache CCD, five Latin-square rotations, 100% minimum pinned-core idle, maximum audited load 1.18, and an explicitly cleared allocator-injection environment.

[codspeed-hq]

Merging this PR will degrade performance by 22.13%

⚡ 4 improved benchmarks
❌ 7 regressed benchmarks
✅ 99 untouched benchmarks

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
❌	JackVec[8]	127.2 ns	278.3 ns	-54.29%
❌	SmallVec4[8]	155.6 ns	307.8 ns	-49.46%
❌	ThinVec[8]	156.1 ns	307.5 ns	-49.23%
❌	SmallVec8[8]	155.6 ns	248.3 ns	-37.36%
❌	Vec[8]	155.6 ns	218.1 ns	-28.66%
❌	ThinVec[4]	132.5 ns	161.7 ns	-18.04%
❌	JackVec[4]	192.5 ns	221.7 ns	-13.16%
⚡	SmallVec4[4]	174.4 ns	145.3 ns	+20.08%
⚡	SmallVec4[4]	598.6 ns	511.1 ns	+17.12%
⚡	ThinVec[1]	1.6 µs	1.4 µs	+10.37%
⚡	ThinVec[4]	1.6 µs	1.4 µs	+10.3%

Tip

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

---

_{Comparing perf/iteration-audit (2d2db35) with main (4943262)}