Feature: Add redis caching for Violation Detection transformation part!

[amindadgar]

As we unwind every document into multiple documents to find separate each activity per author, we have ended up in multiple documents with the same text field. To assure that we're not being costed a high price for violation detection while labeling the data we need to add a caching mechanism to call the LLM API just for the new texts and return the results of previous texts for label.

To do so here's the task list for this issue

• Add caching mechanism for LLM API calling
• Make sure to send reports one per each text and link for users #246