Summary of changes:

• Adds the HumanEval LLM benchmark (@ramkrishna2910)
• Reorganizes the LLM source code into src/lemonade instead of src/turnkeyml/llm (@jeremyfowers)
  • LLM support is now consistently referred to as lemonade instead of turnkeyml-llm throughout the repo
  • BREAKING CHANGE: Code that imported turnkeyml.llm should import lemonade instead
    • Package version number set to v5.0.0 because of this breaking change
• Adds a progress monitor to MMLU accuracy testing (@jeremyfowers)
• Fixes a bug where benchmarking could fail and give a confusing error message (@jeremyfowers)
• Fixes a bug where giving an incorrect MMLU test name would silently fail (@ramkrishna2910)