Add support to LogProbs retrieval by implementing the native Structured Output Parser from OpenAI

[WittmannF]

Is your feature request related to a problem? Please describe.
I've been reading the previous pull requests and found that #916 has been deprecated in favor of #938 which calls structured outputs using tool call by setting strict to True. In theory this is great for compatibility with previous models, however, as reported in #742 , tool/function calls doesn't output logprobs. Here's an example:

import openai
import json

openai_client = openai.OpenAI()

completion = openai_client.chat.completions.create(
    model="gpt-4o-mini",
    tools=[
        {
            "type": "function",
            "function": {
                "name": "ExtractUser",
                "description": "Correctly extracted `ExtractUser` with all the required parameters with correct types",
                "parameters": {
                    "properties": {
                        "name": {"title": "Name", "type": "string"},
                        "age": {"title": "Age", "type": "integer"},
                    },
                    "required": ["age", "name"],
                    "type": "object",
                },
            },
        }
    ],
    tool_choice={"type": "function", "function": {"name": "ExtractUser"}},
    logprobs=True,
    messages=[
        {"role": "user", "content": "Extract Jason is 25 years old"},
    ],
)
print(completion.choices[0].logprobs)

# Output: ChoiceLogprobs(content=None, refusal=None)

While using the native structured output parser from OpenAI will provide logprobs:

import os
from pydantic import BaseModel, field_validator
import openai

client = openai.OpenAI()

class User(BaseModel):
    name: str
    age: int

response = client.beta.chat.completions.parse(
    model="gpt-4o-mini",
    max_tokens=1024,
    messages=[
        {
            "role": "user",
            "content": "Extract John is 18 years old.",
        }
    ],
    response_format=User,
    logprobs=True,
)


response_logprobs = response.choices[0].logprobs.content
print(response_logprobs)

# output: [ChatCompletionTokenLogprob(token='{"', bytes=[123, 34], logprob=-1.9361265e-07, top_logprobs=[]), ChatCompletionTokenLogprob(token='name', bytes=[110, 97, 109, 101], logprob=0.0, top_logprobs=[]), ChatCompletionTokenLogprob(token='":"', bytes=[34, 58, 34], logprob=0.0, top_logprobs=[]), ChatCompletionTokenLogprob(token='John', bytes=[74, 111, 104, 110], logprob=0.0, top_logprobs=[]), ChatCompletionTokenLogprob(token='","', bytes=[34, 44, 34], logprob=0.0, top_logprobs=[]), ChatCompletionTokenLogprob(token='age', bytes=[97, 103, 101], logprob=0.0, top_logprobs=[]), ChatCompletionTokenLogprob(token='":', bytes=[34, 58], logprob=0.0, top_logprobs=[]), ChatCompletionTokenLogprob(token='18', bytes=[49, 56], logprob=0.0, top_logprobs=[]), ChatCompletionTokenLogprob(token='}', bytes=[125], logprob=0.0, top_logprobs=[])]

Describe the solution you'd like
Continue #916

Describe alternatives you've considered
Using the native structured outputs from open ai client (client.beta.chat.completions.parse)