bench audit: rwa-yield-accuracy BUIDL/BENJI + cross-chain/perp-oi disclaimers

[Flotapponnier]

Summary

• Remove BUIDL and BENJI from rwa-yield-accuracy providers — both probes are commented out in main.go (emit no Prometheus metrics), so they were showing as stale/no-data rows
• Update all copy (seo_description, abstract, methodology cohort, findings, FAQ) to reflect the active 4-provider cohort: USDY, USTB, OUSG, SyrupUSDC
• Replace BUIDL-specific FAQ with "Why are BUIDL and BENJI not in the cohort?" explaining the measurement pause for each
• Add disclaimer on cross-chain-messaging-latency about pre-finality vs post-finality security tradeoff
• Add disclaimer on perp-open-interest about gains.trade multi-chain aggregate unfairness vs single-chain venues

[Flotapponnier]

Superseded by #1513 — rwa-yield-accuracy copy fixes merged via that PR.