Improve error message for 401 / invalid API key (avoid "local model" confusion)

[vlabsc]

When using ChatOpenAI with an invalid or wrong API key, the user sees: 401 Invalid model key or Incorrect local model configuration.

This is misleading because:

1. The actual cause was an invalid/wrong API key (401 from the API).
2. The “or Incorrect local model configuration” part suggests Ollama/local setup, which doesn’t apply when using OpenAI.

Suggestion: For 401 responses from the OpenAI (or similar) API, show a clearer message, e.g.: 401 Unauthorized – check your API key and that it has access to the requested model.

Keep a separate message for real local-model issues (e.g. connection to Ollama), so users aren’t sent down the wrong path.

Version: 3.0.13