Parse logs to understand errors

[cliffckerr]

Write an LLM parser for the output logs to understand what questions it's getting wrong and why.

[krosenfeld-IDM]

https://meridianlabs-ai.github.io/inspect_scout/ might be useful here