LLM Eval - Implement custom LLM evaluator component in core

[shadeMe]

Implement a LLMEvaluator component with the following parameters:

• api - API to use (openai, cohere, etc)
• api_key - API secret
• inputs - tuple of expected inputs
  • Will act as the sockets for the component
• outputs - keys to look up in the output JSON
• instruction - prompt to the LLM describing what it should do
  • Prompt should always return a json string
  • We can (try to )ensure this by modifying the prompt internally
• examples - One or more examples conforming to the input and output format

The component sets up the backend, validates/formats the inputs, calls the backend, validates the output and returns them.