Add AgentConversation integration test harness with example project and README

[Copilot]

Motivation and Context

Validates conversation dynamics for long-running, multi-agent, tool-heavy scenarios without rebuilding costly context on every run. The harness captures a representative conversation once, serializes it, and lets subsequent tests restore + validate from that snapshot.

Description

Rename ConversationDynamics → AgentConversation throughout

• Folder, .csproj, all C# namespaces, and solution reference updated
• Old ConversationDynamics.IntegrationTests folder removed

AgentConversation.IntegrationTests (base class library, IsTestProject=false)

Core harness abstractions:

Type	Role
IConversationTestCase	Agents, initial messages, ordered steps, automated context-creation method
IConversationTestSystem	Plugin for agent construction and optional compaction — both vary per backend
ConversationAgentDefinition	Name, instructions, tools per agent
ConversationStep	Agent + optional input + optional Action<AgentResponse, ConversationMetricsReport> validator
ConversationContextSerializer	JSON serialize/deserialize IList<ChatMessage> to/from files
ConversationHarness	Restores context into InMemoryChatHistoryProvider, compacts if requested, executes steps, captures before/after metrics
ConversationHarnessTests<TSystem>	Abstract xunit base; inherits RunAllTestCasesAsync and SerializeAllInitialContextsAsync (skipped by default)

Microsoft.Agents.AI.Abstractions.IntegrationTests (runnable test project)

Self-contained example that exercises the harness with a mock IChatClient — no live AI credentials required:

// Concrete test class — inherit and wire up system + cases
public class AgentConversationHarnessTests(ITestOutputHelper output)
    : ConversationHarnessTests<InMemoryConversationTestSystem>(output)
{
    protected override InMemoryConversationTestSystem CreateTestSystem() => new();
    protected override IEnumerable<IConversationTestCase> GetTestCases() =>
        [new MenuConversationTestCase()];
}

• InMemoryConversationTestSystem — IConversationTestSystem backed by Moq; shows how to wire a real backend
• MenuConversationTestCase — IConversationTestCase with inline messages, restaurant menu tools, and a validation step
• References both AgentConversation.IntegrationTests and Microsoft.Agents.AI.Abstractions

AgentConversation.IntegrationTests/README.md

Documents all abstractions, step-by-step usage, context fixture generation workflow, and metrics output format.

Contribution Checklist

• The code builds clean without any errors or warnings
• The PR follows the Contribution Guidelines
• All unit tests pass, and I have added new tests where possible
• Is this a breaking change? No

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.