Show selectable model comparisons and keep local app current

[mattfaltyn]

Summary

• Update a clean local main to canonical origin/main before locked installs; report offline, newer remote, and frontend/backend mismatch states.
• Show five physical forecast models and two separate risk-neutral quote models on the chain and watchlist. The selected physical model drives compact odds, sorting, and hypothetical risk without changing forecast issuance or promotion state.
• Display dated, provenance-separated model evidence and exact unavailable reasons, with bounded background fitting and refresh.

Verification

• Scoped Pad check: managed.
• scripts/verify-fast: 407 backend and 149 frontend tests passed.
• scripts/verify: Ruff, 407 backend and 149 frontend tests, npm audit, frontend lint, generated API check, build, and 27 Playwright tests passed.
• Independent R2 and security reviews found no remaining actionable findings.

Current prospective evidence is descriptive; selecting an experimental model does not promote it or claim superior accuracy.