create an example of how to improve a retriever on industry dataset

[mrm1001]

Goal:
Showcase the Haystack evaluation metrics on an example that is close to what our users are trying to do, i.e. to improve a retriever in a RAG app.

The goal of this task is to find such an example, ideally finding:

• the RAG pipeline architecture
• the metrics to be used (if Haystack metrics are not finished yet, can use ragas/deepeval implementations for now). Here is the list of metrics that will be released in Haystack
• the experiment to show how the retriever improves and the improvement should show in the metrics: ideally after tweaking chunk size, or changing embedding model/size.

This does not need to be a finished tutorial or even code, just need some scrap code somewhere (colab?).

Depends on: #7438