Add vllm library to synda

[comeRodriguez]

This PR try to add vLLM as a possible provider for batch GENERATION step in pipeline

Features

• Added vLLM programmatic API and rest API call to generate new data
• Created a VLLMProvider class to handle vLLM calls

Warnings
Even if PR is ready to review, work is still in progress (no vLLM batch inference for ABLATION step). Any help or advice is more than welcome :)