v0.2.3

What's new

• Parallel pricing fetch + 24h disk cache: startup drops from ~5s to ~2s on first run, ~1s on subsequent runs
• Pricing prompt retry: bad input on any price field re-asks that field only; Ctrl+C still skips the model
• Total Tokens column: shown when prompt cache is inactive, giving an at-a-glance input + output sum
• Bug fix: cleanup_overrides no longer throws a TypeError when pricing is loaded from cache