perf(ci): surface keyed-list allocation in the /perf comment

[azchohfi]

What

The keyed-list /perf leg (StressPerf.KeyedList) already captures the full allocation set in metrics.json (allocBytesPerRender, gen0PerKRenders) via the shared PerfTracker, but Format-PerfKeyedListSection rendered only the four headline metrics — so the keyed-diff allocation signal was invisible in the comment.

This adds an Allocation (keyed-list) sub-table inside the keyed-list section, reusing the existing $PerfAllocMetricSpec over the keyed aggregates with the identical paired-Δ 95% CI machinery as the headline and StocksGrid allocation tables.

Allocation on the keyed (ReconcileKeyed → ReconcileKeyedMiddle / LIS) path scales with reorder volume in a way the positional StocksGrid allocation table can never isolate — the exact signal keyed-diff alloc-reduction work (e.g. #657) needs measured.

Why it's safe

• Pure render reuse — no workload, harness, or metrics.json change. The keyed aggregates already carry AllocBytesPerRender / Gen0PerKRenders (+ Spread / Samples) via Measure-PerfRuns, exactly like the StocksGrid leg.
• Gated on the keyed leg reporting alloc metrics (n/a for a legacy head), mirroring the StocksGrid alloc table's $hasAlloc guard.
• Footprint: 3 files, all tests/stress_perf/ci/**, 0 src/Reactor (Class-B harness-only).

Rendered output

Below the keyed headline metrics, the section now shows (real magnitudes from a StressPerf.KeyedList @50% run, alloc reduced ~20%):

Allocation (keyed-list) — lower is better

Metric	main	This PR	Δ (95% CI)	Status
Alloc bytes/render ↓	328000	262000	-20.1%	✅ improvement
Gen0 GC / 1k renders ↓	63.20	50.20	-20.6%	✅ improvement

Tests

+8 assertions in PerfLib.Tests.ps1: sub-table renders, bytes/render + Gen0 rows present, alloc-down reads improvement, ordering after the headline table, omitted when keyed aggregates lack alloc, and distinct from the StocksGrid Allocation (Reactor) table in a full comment. PerfLib.Tests 226 green; RunPerfBenchmark.Tests 39 green.

---

Part of the "make /perf valuable" harness roadmap (follows PR1–PR4 + #700). Class-B harness-only.

[Copilot]

Pull request overview

Surfaces allocation metrics for the keyed-list /perf leg (StressPerf.KeyedList) in the rendered PR comment by adding an allocation sub-table to the keyed-list section, using the existing allocation metric spec and the same paired-Δ 95% CI machinery already used elsewhere in the perf comment.

Changes:

• Extend Format-PerfKeyedListSection to append an Allocation (keyed-list) markdown table (Alloc bytes/render, Gen0 GC / 1k renders) when allocation metrics are present.
• Add PowerShell unit tests to verify the keyed allocation sub-table renders/omits correctly and remains distinct from the StocksGrid allocation table.
• Update the perf CI README to document the new keyed allocation sub-table and its intent.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
tests/stress_perf/ci/PerfLib.ps1	Adds a keyed-list allocation sub-table to the keyed-list section, gated on alloc metrics being present.
tests/stress_perf/ci/PerfLib.Tests.ps1	Adds assertions covering rendering, ordering, omission gating, and distinction from the StocksGrid allocation table.
tests/stress_perf/ci/README.md	Documents that keyed-list output includes its own allocation sub-table and when it appears.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated no new comments.