[WIP] Create integration test harness for conversation dynamics

[Copilot]

Thanks for asking me to work on this. I will get started on it and keep this PR's description up to date as I form a plan and make progress.

Create a test harness as an integration test under dotnet/tests. This test harness is used to validate conversation dynamics for long running operations based on multiple agents that are using tools. Instead of building up the long context everytime, let's start by deserializing an array of chat messages. From that restored context, one or more agent responses can be solicited and validated. The before and after metrics of the conversation should be captured and reported on. Each context inherently involves one or more agents so they need to be defined for each case as well as the inputs and expectations. The specific manner in which each agent is created can vary, based on the system being tested, so make that easily adapt to different systems. The specific contract for performing compaction may also vary across systems, so that also should be dynamic. Finally, each test case must support the ability to create and serialize the initial context. Even though this initial setup doesn't need to occur for each test, it does need to be automated.

---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.