Bench: add LLM judge for response scoring to chat strategy

[abeatrix]

• Add llmJudgeChatTemplate to generate a prompt for evaluating LLM responses
• Integrate LlmJudge into the evaluateChatStrategy to score each chat response
• Store the numeric score in the EvaluationDocument for each chat response
• Log the total score for the fixture at the end of the evaluation

There are still follow-up works that we can do, including improving the prompt used for the LLM judge, but currently, it works as intended and IMO a good starting point for us to build from.

Test plan

Currently being used by https://github.com/sourcegraph/cody-leaderboard/pull/8

[github-actions[bot]]

‼️ Hey @sourcegraph/cody-security, please review this PR carefully as it introduces the usage of an unsafe_ function or abuses PromptString.

[abeatrix]

as discussed with @jtibshirani , i am running into issue with re-running the tests so we will merge what we have now and update the prompt and scoring in follow up PRs