Feature: support default model and effort settings in /codex:setup

Add support for persisting default --model and --effort values via /codex:setup, so users don't have to pass them on every command invocation.

Motivation

Currently, --model and --effort must be specified per-command. Users who consistently prefer a specific model (e.g., gpt-5.4) or effort level (e.g., xhigh) have to repeat these flags every time they run /codex:rescue, /codex:review, or /codex:adversarial-review.

The setup command already persists state like the review gate toggle, so it's a natural place for these defaults.

Proposed UX

/codex:setup --default-model gpt-5.4 --default-effort xhigh

• Defaults would apply to all commands (task, review, adversarial-review) unless overridden per-invocation with explicit --model or --effort flags.
• /codex:setup (without flags) and /codex:status would display the current defaults.

Additional context

• --effort is currently only supported on task (rescue), not on review or adversarial-review. It may make sense to extend --effort support to review commands as part of this work.
• Claude Code's /fast mode toggle could also be worth supporting.

Start by tracing how /codex:setup persists the review gate and how /codex:status reads it, then follow option handling for task, review, and adversarial-review. Done means setup can store and display model and effort defaults, commands use them unless explicitly overridden, and the supported scope for review effort is resolved.