For OpenAI and Azure we should allow any LLM to use any embedded. We do not need to change the onboarding flow or other screens and instead:

When using LLM = openai, assume OpenAi is the embedder
When using LLM = Azure, assume Azure is the embedder

We should just enable the user to use another embedder if they prefer.

Onboarding data handling modal should now show embedder if the LLM is one that does not use a built-in embedding engine.