fix(codex): 修复 Codex provider 同步功能

[Fldicoahkiin]

根据审查意见修复 Codex provider 同步功能：

修复内容

1. sync_codex_rollout_model_provider: 只修复缺失值（空字符串），不覆盖已有值
2. sync_codex_threads_model_provider: 只修复 NULL 值，不覆盖已有 provider 记录
3. 删除 rebuild_codex_session_index: 基于错误前提（threads 表不是 session_index 的全量来源）
4. 执行路径优化: 只在 switch_normal 中执行，不在热切换路径执行（避免重 I/O 操作增加切换耗时）
5. 添加警告注释: 明确竞态条件风险

变更统计

• src-tauri/src/codex_config.rs: 修改同步策略，删除重建函数
• src-tauri/src/services/provider/mod.rs: 删除重建函数调用，优化注释

Closes #1694

[Fldicoahkiin]

已根据审查意见修复：

1. sync_codex_rollout_model_provider: 只修复缺失值（空字符串），不覆盖已有 provider 记录
2. sync_codex_threads_model_provider: 只修复 NULL 值，不覆盖已有值
3. 删除 rebuild_codex_session_index: 基于错误前提（threads 表不是 session_index 全量来源）
4. 执行路径优化: 只在 switch_normal 执行，不在热切换路径执行

请 review。

[farion1231]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 25b38368d8

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Use Codex model_provider key for metadata sync

The id passed here is the cc-switch provider record ID, but Codex session metadata is keyed by TOML model_provider values (for example custom/openai), not by internal UUIDs. In this repo, non-additive providers are created with generated UUID IDs, so using id causes both sync helpers to write a value that does not match Codex’s active provider key, which means thread/list filtering can still miss or misclassify history after switching. Please parse the target provider’s settings_config.config and pass its model_provider value to these sync functions instead of the internal provider ID.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Avoid full-file rewrite during rollout metadata patch

This path rewrites the entire JSONL file from an in-memory snapshot after reading it, so if Codex is concurrently appending events to the same rollout file during a switch, any lines appended between the read and the atomic replace are lost. The severity depends on concurrent writes, but that scenario is realistic for active sessions and leads to irreversible history truncation. The sync should avoid replacing whole files (or use locking/active-session exclusion) when only the first line needs patching.

Useful? React with 👍 / 👎.