Performance optimization & benchmark suite

[kafkade]

Tracks Phase 5.2 — Performance Optimization from docs/SPEC.md. This is the only Phase 5 feature from the spec without a tracking issue.

Goal

Production-grade performance with regression protection, meeting the non-functional requirements in SPEC §5.

Scope

• Add a criterion benchmark suite (benches/) covering hot paths (collectors, ring buffer, rendering)
• Efficient diff-based rendering — only redraw changed cells
• Lazy collection — only collect data for visible panels
• Wire benchmarks into CI for performance-regression detection

Targets (from SPEC §5 NFRs)

• CPU overhead: < 1% at 1s refresh on modern hardware
• Memory: < 30MB RSS typical, < 50MB with all features
• Startup: < 200ms to first render
• Binary size: < 15MB stripped (static Linux musl build)

Notes

• No benches/ directory or criterion dev-dependency exists yet.
• Lazy collection touches the collector → App → UI data flow; per SPEC §7, prefer tokio::sync::watch over Arc<RwLock<>> if channels are introduced.