Evaluation engine may not respect choice of model

[caufieldjh]

Running an evaluation may ignore the supplied model and use the default model (gpt-3.5-turbo) instead.
With the following, for example:

ontogpt -vvv eval -o soup.yaml --num-tests 5 --no-chunking -m gpt-4 EvalCTD

The client appears to still be using the default:

INFO:ontogpt.clients.openai_client:Complete: engine=gpt-3.5-turbo, prompt[769]=Split the following piece of text into fields in the following format:

BUT...it also is using a cached response, so either way the specified model isn't used.

[caufieldjh]

Hrm, even with a fresh cache it isn't passing the value for the model param.