Fixes for openai server example

[mgoin]

Thanks to @dsikka for noticing the current outputs of the server were garbage - the pipeline was returning only the tokens delta during streaming rather than the ongoing concatenated text.

Also implement full support for non-streaming.

Start the server with:

python examples/openai-server/server.py --model zoo:nlg/text_generation/codegen_mono-350m/pytorch/huggingface/bigpython_bigquery_thepile/base-none

And hit it with a prompt request:

curl http://localhost:8000/v1/completions \                                                                                                                                  
    -H "Content-Type: application/json" \
    -d '{
        "model": "zoo:nlg/text_generation/codegen_mono-350m/pytorch/huggingface/bigpython_bigquery_thepile/base-none",
        "prompt": "def fib():",
        "max_tokens": 30
    }'  
    
{"id":"cmpl-90fc6217001d4556973b2b82ebc44270","object":"text_completion","created":1693451970,"model":"zoo:nlg/text_generation/codegen_mono-350m/pytorch/huggingface/bigpython_bigquery_thepile/base-none","choices":[{"index":0,"text":"\n    a, b = 0, 1\n    while True:\n        yield a\n        a, b = b, a + b","logprobs":null,"finish_reason":"stop"}]}