perf: establish release-mode workload, allocation and idle-resource baselines

[subinium]

Priority: P2
Evidence: Existing debug measurements only; release qualification work
Baseline: v0.23.0 source 37933f9, audited 2026-09-05. Native execution/compile probes used Rust 1.98.0 on aarch64 macOS unless explicitly stated otherwise.

Problem and Reproduction

The audit reproduced fixed frame allocations and quadratic scroll discovery, but most remaining speed/RSS claims are not measured. Existing checks cover many features and OSes; they do not establish the effect of future optimizations across realistic TUI workloads.

benches/benchmarks.rs:1 · tests/v020_perf_alloc.rs:94 · .github/workflows/ci.yml:185

Scope

• Add repeatable release baselines for idle, typing/paste, large filtered collections, resize, static logs, charts and image updates.
• Record toolchain, feature set, host, geometry, data size, warm-up and sample method.
• Distinguish CPU/kernel/flush/wall time, allocator calls/requested bytes and retained/peak memory.

Expected Effect

Choose optimizations by demonstrated user-visible cost and prevent unsupported performance claims.

Acceptance Criteria

• Report p50/p95 frame/input latency, idle CPU, allocations/bytes per frame, output bytes and peak/retained bytes.
• Use80x24,120x40,240x80 plus repeated resize; fixed deterministic data fixtures.
• Do not promote debug closure timings to release full-frame throughput.
• Gate deterministic allocation/correctness regressions; avoid flaky shared-runner nanosecond thresholds.
• Run current patched stable and MSRV1.88 where supported; keep the declared MSRV unless an explicit compatibility decision changes it.
• Follow the repository quality gates and the evidence rules in tracking: post-v0.23 correctness, resource efficiency and WASM completion #366; record actual results rather than treating a proposed test as passed.

Boundaries and Dependencies

Current CI already covers Linux/macOS/Windows and real multiplexers. This adds targeted measurements and semantic assertions, not a replacement CI framework.

No blocking implementation dependency identified; coordinate shared code with the tracker.

Tracking: #366. This issue records unresolved work, not an implemented fix.

[subinium]

v0.24.0 Disposition

Release diagnostics and docs/PERFORMANCE.md now record toolchain, workload geometry, allocation calls/requested bytes, retained storage, command counts and bounded latency samples. Noisy or regressed measurements are retained. Missing broader acceptance includes long-duration terminal idle CPU/output bytes, full image transport and comprehensive input-to-display p95. Requested allocator bytes are not RSS, and full-suite RSS is not a per-widget retained-memory measurement.

Implementation and evidence mapping: PR #433. Release v0.24.0 and its 20/20 release workflow are verified. This issue stays open only for the remaining acceptance/investigation described above; publication alone does not establish it.