[Measurement] Sort findingMetricsByType keys in benchmark runner for stable baseline diffs

[AndresL230]

Summary

benchmark/baseline.json's findingMetricsByType object preserves insertion order from the runner, which follows actual-emission order. This causes cosmetic diff noise across PRs whenever the set of emitted finding types changes (PR-2 removed cache, PR-4 removed rate_limit — each refresh reshuffles the remaining rows).

Evidence

• PR C1 PR-4: tighten rate_limit and residual batch FPs (closes #83) #111 baseline diff included a row-reordering chunk in addition to the actual value changes.
• A reviewer pointed this out as Minor in the final code review.

What to do

• In benchmark/runner.ts main() (around the --update-baseline write), sort the keys of findingMetricsByType alphabetically before serialization. Same for any other dynamic-key map in the JSON.
• Pure runner change; does not touch the metrics computation.

Acceptance criteria

• findingMetricsByType keys appear in alphabetical order in baseline.json after --update-baseline.
• Refresh baseline once to apply the sort; subsequent PR diffs touch only value changes, not order.
• No metric values change as a result.

Reference

Final code review on PR #111.