fix(dashboard): separate live and published context ratios

[AjnasNB]

What changed

• keep the real per-workspace comparison and label it This workspace
• add a separate 77.81:1 published six-fixture estimate card with 442,113 → 5,682 estimated tokens and 98.71% less repeated context
• include both scopes in the memory-footprint table and public dashboard types/docs
• bind the published values to bench/results/benchmark-release-0.1.6.json in regression tests
• regenerate the committed Codex and Claude plugin runtimes

Root cause

The dashboard showed only the dynamic workspace ratio. For the current ledger that is correctly 64.96:1, but without the published benchmark beside it the UI made it easy to mistake that workspace-specific value for the published 77.81:1 six-fixture result.

Validation

• live loopback page: 64.96:1 and 77.81:1 both present with distinct scope labels
• focused dashboard/platform/memory-footprint tests: 13/13 pass
• full npm run check: pass in 307.1s
• site: 34 pages built and validated
• MCP smoke, TypeScript, benchmark/research evidence, deterministic evaluations, plugin sync, and package dry-run: pass

[AjnasNB]

All 13 exact-head checks are green, including the repaired Ubuntu Node 26 status and CodeQL. Local full npm run check passed in 307.1s. Ready for current-head review; auto-merge is armed.