bench: statistical methodology (warmup, confidence intervals)

[danieljohnmorris]

Summary

Implements ILO-349 (follow-up to ILO-65 / #608).

• Warmup: 5 runs discarded before measurement (3 in --quick mode) to eliminate cold-start noise
• 30 measurement runs per engine/language (10 in --quick), each an independent process invocation; ilo uses its internal 10k-iter loop per run
• Full statistics computed per (benchmark, language): min, max, mean, median, p95, p99, stddev
• 95% confidence intervals via t-distribution (df = n-1), converging to z = 1.960 for large samples
• CI overlap detection: after all benchmarks complete, all pairwise CIs are compared; overlapping CIs are reported as "no statistically significant difference" (non-fatal warning, exit 1 can be enabled for CI gates)
• results.json schema updated: values are now stats objects instead of bare scalars; a methodology block records warmup/measure counts and CI method

Test plan

• ./bench/run.sh --quick --no-rust completes without error
• bench/results.json contains ci95_lo/ci95_hi fields
• CI overlap section prints without crashing on degenerate (single-sample) input
• --quick produces 10 samples, normal mode produces 30

🤖 Generated with Claude Code

[codecov]

❌ 1 Tests Failed:

Tests completed	Failed	Passed	Skipped
686	1	685	0

View the full list of 1 ❄️ flaky test(s)

ilo::interpreter::tests::interpret_braced_guard_in_loop_no_early_return

Flake rate in main: 66.67% (Passed 1 times, Failed 2 times)

Stack Traces | 0.011s run time

thread 'interpreter::tests::interpret_braced_guard_in_loop_no_early_return' (32872) panicked at src/interpreter/mod.rs:10202:9:
parse errors: [ParseError { code: "ILO-P003", position: 27, span: Span { start: 35, end: 36 }, message: "expected `{`, got `;`", hint: Some("ilo bodies are single-line, `;`-separated — not python/swift-style indented. Use either the brace-block form `name p:t>r;{body1;body2}` or the single-line form `name p:t>r;body1;body2`. For statements that require a block (`@k xs{...}`, `wh cond{...}`, `?subj{...}`), the `{...}` must be on the same line as the head.") }]
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

To view more test analytics, go to the Test Analytics Dashboard
_{📋 Got 3 mins? Take this short survey to help us improve Test Analytics.}

[danieljohnmorris]

needs manual rebase (conflicts in: bench/run.sh)

[danieljohnmorris]

needs manual — rebase conflict in non-doc file(s). Auto-resolve skipped.

[danieljohnmorris]

needs manual — rebase conflict in non-doc file(s)