enhancement: collapse perfect-parity audits into a single bot-summary line

[BraedenBDev]

Follow-up from docs/issues/2026-04-12-accuracy-from-live-audit.md (R3) and docs/plans/2026-04-18-live-audit-follow-through.md.

Problem

crawl-sim now computes parity, but the user-facing presentation still has a lingering ambiguity: when all bots effectively see the same thing, redundant per-bot rows add noise and imply differentiation where there is none.

Goal

Make parity-first presentation the default when bot deltas are below a meaningful threshold, while preserving the full per-bot breakdown when divergence is real.

Scope

• Define the collapse threshold for user-facing output/report rendering
  • likely something like max(score) - min(score) <= 5 and near-identical visibility/parity metrics
• Collapse the scorecard/report into a single summary line when that threshold is met
• Keep the expanded per-bot table when divergence is meaningful
• Ensure HTML/PDF/report renderers and SKILL examples stay consistent with the new behavior

Acceptance criteria

• Perfect or near-perfect parity audits render as a single summary line
• Divergent audits still render a full per-bot breakdown
• Threshold logic is tested with one parity fixture and one divergence fixture
• README/SKILL/report examples reflect the parity-collapse behavior

Why this matters

The live audit's “four rows of 94/A” were technically true but semantically noisy. Parity should be legible at a glance.