Modify benchmarking tests and set up github workflows to auto-run benchmarks on any PRs

[danXyu]

• Requires setting up good testing benchmarks for each of the agents that run as pytests.
• Then we can set up a github workflow to automatically run these benchmarks on PRs.
• Will ensure that any changes we make, whether that's on the prompts, the LLM model, or anything, don't affect the overall experience / the LLM can still delegate to the appropriate agents.