Implement capability to compare response against different LLM

[edwardmfho]

Currently promptimize only evaluate results as pre-defined by developers. We could potentially leverage the current tool, to compare response against different LLMs. For example, we could use GPT-4 as a benchmark, to evaluate the responses from a custom model, on whether they produce similar results.

• Add in optional parameter for defining a "target" LLMs.
• Add in function to compare the "similiarity" of the results
• Allow user to still utilise manual test cases to compare the two LLMs being evaluated.