Feature request: import/sync sessions from Codex, mirroring ChatGPT's existing "import from Claude Code" feature

## Summary

OpenAI's ChatGPT/Codex app has a native, built-in feature (Settings → Import → "Importar desde otra aplicación de IA") that detects local Claude Code session data and imports it into ChatGPT, with an "auto-sync" toggle ("Mantener las importaciones sincronizadas") that keeps pulling in new Claude Code sessions and projects automatically over time. It explicitly lists "Claude Code" as a supported import source, alongside project folders and chat sessions (a real-world example: 17 projects / 42 chat sessions offered for import in one pass).

Claude Code has no equivalent feature in the other direction — there is no way to import or view Codex/ChatGPT sessions from within Claude Code.

## Why this matters

Users who work across both tools (e.g. running some tasks in Claude Code and others in Codex on the same projects, or switching to Codex when Claude usage/quota runs out and vice versa) currently have a one-directional bridge: their Claude Code work flows into ChatGPT automatically, but their Codex/ChatGPT work is invisible from Claude Code. This makes it hard to keep a single coordinated view of work spread across both tools, and effectively nudges users toward consolidating their session history inside the competitor's app rather than Claude Code's own interface.

Today the only workaround is fully manual: asking Claude to read a specific Codex session's local rollout file (`~/.codex/sessions/YYYY/MM/DD/rollout-*.jsonl`) on request and summarize it in the current conversation — there's no persistent, synced, or automatic visibility.

## Requested behavior

Add an equivalent "Import from another AI app" capability in Claude Code (CLI and/or desktop app settings) that can detect and import local Codex/ChatGPT session data, ideally with:
- A one-time import of existing Codex projects/sessions.
- An optional auto-sync toggle to keep pulling in new Codex sessions over time, mirroring what ChatGPT already offers for Claude Code.
- Some way to distinguish imported (Codex-origin) sessions from native Claude Code ones in the session list/sidebar, so users don't have to manually prefix titles themselves to tell them apart.

## Environment

- macOS (Darwin)
- Observed via ChatGPT/Codex desktop app's own Settings → Import page, which explicitly lists "Claude Code" as a ready-to-import source with auto-sync already enabled.

The issue identifies ~/.codex/sessions/YYYY/MM/DD/rollout-*.jsonl as the source and mentions Claude Code's CLI and desktop settings as entry points; begin by locating session discovery, import, and session-list code in this repository. No specific files or tests are named, so first map those areas and any existing import behavior. Done means one-time Codex session/project import, optional synchronization, and a visible distinction between imported and native sessions.

Written by the indexing model from the issue text.