FEAT: Adding Converters to Output

[rlundeen2]

Description

• Adds Converters to the output. This enables several obvious scenarios, such as, if your responses are a certain way, you can convert them (e.g. if the LLM responds in base64 encoded text, you can base64 decode it with a converter).
• Removing synchronous send_prompt methods
• Moving memory additions from targets to the normalizer
  • This reduces code to add memory for every target
  • This also simplifies exception handling - the behavior is the same but there is less code to add exceptions to memory
• Moved scoring docs to their own folder for better readability and easier notebook generation
• Minor scoring improvements because notebooks were failing consistently with the asserts

Tests and Documentation

Unit tests added. Jupytext passes