Using DISABLE_MODEL_SERVER=false causes crash on load

[jamesyc]

Repro steps:

1. create docker containers via quick start script, or manually copy docker-compose.yml
2. edit the .env file to include DISABLE_MODEL_SERVER=True (or delete the inference_model_server service from the docker compose file)
3. start the docker containers with docker compose start -d

Expected behavior:
Onyx api and web server starts correctly without using inference_model_server

Actual behavior:
Onyx api crashes at search_nlp_models.py line 1069

Suggested fix:
Modify search_nlp_models.py to respect DISABLE_MODEL_SERVER flag so that the app runs with model server off