[Bug]: Retry remaining batch of prompts on failure

[yadavsahil197]

Describe the bug
The labeling agent sends CHUNK_SIZE no of prompts for labeling to the LLM at a given time. If the LLM fails to label some of the prompts, the entire chunk of prompts are left unlabeled.

To Reproduce
A list of steps to reproduce the behavior:
Use an LLM with smaller context window and create a prompt example with large enough tokens, all prompts in the batch fail even if some of them have tokens within the context window of the model.

Expected behavior
The labeling agent should fail only the actual failure prompt and retry for the remaining prompts in the batch.

[Abhinav-Naikawadi]

@Tyrest we can close this right? Done here: #443