Add logprob support for OpenAI models

[tom-doerr]

This is part of #879.
Is this way of adding support for logprobs acceptable?

lm = dspy.OpenAI(model='gpt-3.5-turbo-instruct', max_tokens=6, api_key=config['openai']['secret_key'])
test_text = "This is a test article."
test_output = lm(test_text, logprobs=1)
print("test_output:", test_output)

test_output: [{'text': '\n\nLorem ipsum dolor sit amet', 'logprobs': {'text_offset': [23, 25, 30, 36, 42, 46], 'token_logprobs': [-0.26951057, -0.28366262, -0.0010195904, -5.4312077e-05, -0.0001788689, -0.00023739056], 'tokens': ['\n\n', 'Lorem', ' ipsum', ' dolor', ' sit', ' amet'], 'top_logprobs': [{'\n\n': -0.26951057, '\n': -2.4457762}, {'Lorem': -0.28366262, 'Test': -2.9440362}, {' ipsum': -0.0010195904, ' Ipsum': -6.9318686}, {' dolor': -5.4312077e-05, ' d': -11.2004175}, {' sit': -0.0001788689, ' amet': -9.340998}, {' amet': -0.00023739056, ' am': -8.778377}]}}]

If this is okay I would look into adding it to dspy.Predict and the vllm backend.

[arnavsinghvi11]

Thanks @tom-doerr !