ci: land lib-cmp comparison-bench workflows on main (fidelity + perf)

[jackmoxley]

Lands the two comparison-bench workflows on the default branch so they become UI-dispatchable (the workflow_dispatch 'Run workflow' control only appears for workflows on the default branch) and so the pull_request->main hook fires (GitHub reads dispatch/PR workflow definitions from the default branch).

• lib-cmp-precision.yml — the Fidelity / LSBε precision shootout (decimal-scaled vs 6 peers; per-cell precision-relative grade, score + %CR two-letter grade, report to the run summary + artifact).
• lib-cmp-perf.yml — the peer-crate timing comparison (width × scale).

Both are advisory — never required status checks. Trigger: pull_request -> main + workflow_dispatch. They run against the selected ref's tree (release/0.5.0 carries the lib_cmp benches today; they arrive on main with the 0.5.0 release), so on main they stay dormant-but-valid until then. No source changes — workflow files only.

[jackmoxley]

looking good

[codspeed-hq]

Merging this PR will not alter performance

⚡ 4 improved benchmarks
❌ 5 regressed benchmarks
✅ 201 untouched benchmarks
⏩ 9 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
❌	div_ceil	383.6 ns	441.9 ns	-13.2%
⚡	ceil	318.1 ns	288.9 ns	+10.1%
❌	trunc	316.7 ns	375 ns	-15.56%
❌	next_power_of_two	62.2 ns	91.4 ns	-31.91%
⚡	clamp	156.7 ns	127.5 ns	+22.88%
⚡	TryFrom_u128	125.6 ns	96.4 ns	+30.26%
❌	to_num_i32	288.3 ns	346.7 ns	-16.83%
❌	to_f32	220.3 ns	249.4 ns	-11.69%
⚡	D38_div	791.9 ns	675.3 ns	+17.28%

Tip

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

---

_{Comparing ci/lib-cmp-workflows (6ad9994) with main (a9c0e9b)}

Footnotes

1. 9 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. ↩