feat: added langchains LLM

[jjmachan]

fixes #46
fixes #53

You can now use langchains LLM abstraction to access all the LLM endpoints langchain supports.

eg

from langchain.chat_models import ChatOpenAI

gpt4 = ChatOpenAI(model_name="gpt-4")
gpt4.generate_prompt(prompts=[prompts])

# init a new Metric with llm
cr = ContextRelevancy(llm=gpt4)
cr.init_model()
result = cr.score(ds.select(range(4)))

result["context_relavency"]
# [0.46687018871307373, 0.1532887363433838,0.17359847468989234, 0.17340516530234237]

We're also now using OpenAI's chat models as default which brings a 10x decrease in cost. also using gpt-3.5-turbo-16k as default for even bigger context size