Dynamically load Tokenizer after selecting Embedding Engine

This could use a review please from anyone who isn't using an openai embedding model to ensure that it A) maintains what you had selected and B) let's you resave the server and maintains what you had selected. Had fun with subform states and not sure its right...

Screenshot so its easier to see what to test:

Require review from someone not using OpenAI.

Is this still something we want? It has merge conflicts, but if it gets updated I can look into it right away.

Yeah I still think it's useful - the less decisions a user has to make the better

I updated this branch and pointed it to 1.1.x. - this is reviewable, correct @scott_euser?