docs: correct docs for answer relevancy (#86)

Author: jjmachan

docs/metrics.md

@@ -35,7 +35,7 @@ results = context_rel.score(dataset)
 This measures how relevant is the generated answer to the prompt. If the generated answer is incomplete or contains redundant information the score will be low. This is quantified by working out the chance of an LLM generating the given question using the generated answer. Values range (0,1), higher the better.
 ```python
 from ragas.metrics.answer_relevancy import AnswerRelevancy
-answer_relevancy = AnswerRelevancy(model_name="t5-small")
+answer_relevancy = AnswerRelevancy()
 # Dataset({
 #     features: ['question','answer'],
 #     num_rows: 25