Added rag_eval.py

[hungchiayu1]

I modified the way Trulen does rag eval. Previously for groundedness score, Trulens will call TruLlama.select_source_nodes().node.text.collect() which I assume is to select nodes and collect context chunks into a list and pass it into the groundedness feedback function. Right now, I simply pass in the retrieved text into the groundedness feedback function. This should work ( i guess)

[Dhanushkmr]

ill test first then merge later thanks