v1.0.1 — False-positive fixes + Trust button

• Added "Trust this site (don't warn me again)" button to heuristic-warning
modal — your trusts persist across sessions, fully local
• Expanded known-legit-brand whitelist (suno.com, udio.com, krea.ai, claude.ai,
mistral.ai, vercel.com, replit.com, and 30+ other short-domain brands that
previously triggered false typo-squat warnings)
• No changes to phishing-block functionality — still 144k+ domains, daily
refresh, fully local