fix: 400 error when using other providers after logging into OpenAI

[Dcatfly]

If the user is logged into openai, i.e., the ~/.codex/auth.json file exists, and the user switches to other providers, the openai key is still used, which will cause a request error. Related issues: #987

This modification mainly includes the following points:

1. Fix the logic of reading the key, and change it to reading the key based on the provider
2. When ~/.codex/auth.json does not exist:

• Currently: Regardless of the current provider, whether the corresponding key exists in the env, it will display "Login to ChatGPT" and "Enter OpenAI key".
• After the change:
  • If the current provider is OpenAI, it will display 1) Login to ChatGPT 2) Manually enter OpenAI key.
  • If the current provider is not OpenAI
    • If the key corresponding to the provider exists in the env, it is used directly.
    • If it does not exist in the env, only manually enter provider key.

3. Optimized the logic for obtaining savedTokens, abstracting it into the loadAuth method.
4. Optimized the assignment logic for process.env["OPENAI_API_KEY"] to prioritize fetching the value from process.env; if it's not there, it will fetch from auth.json.

[github-actions]

All contributors have signed the CLA ✍️ ✅
_{Posted by the CLA Assistant Lite bot.}

[Dcatfly]

I have read the CLA Document and I hereby sign the CLA