LlmInputs - Updated API + generic framework + convert to vLLM format

[nv-braf]

I've updated the API, added a generic JSON framework to store the intermediate data and added methods to convert to vLLM format.

Here's an example output of HF input to vLLM output. Both role text strings (system/user) are concatenated to form the text input.

{
  "data": [
    {
      "text_input": [
        "You are an AI assistant. You will be given a task. You must generate a detailed and long answer.",
        "Generate an approximately fifteen-word sentence that describes all this data: Midsummer House eatType restaurant; Midsummer House food Chinese; Midsummer House priceRange moderate; Midsummer House customer rating 3 out of 5; Midsummer House near All Bar One"
      ],
      "stream": [
        true
      ]
    },
]
}

[debermudez]

had to remove the "./" in my branch. You will see a slight merge conflict without that update.

[nv-braf]

Fixed to match and am now using this define inside the class

[debermudez]

nice!