Add research note: Microsoft Agent Framework

[rkoster]

Adds a research note on Microsoft Agent Framework (github.com/microsoft/agent-framework) for the agentic runtime research phase. See IDEATION.md for process context.

[rkoster]

Correction (2026-08-10): the original version of this note claimed no Dapr integration exists in Microsoft Agent Framework. That was wrong — I missed the Agent_With_Dapr sample, which uses Dapr's Conversation API as a pluggable inference-backend provider, and Diagrid's dotnet-ai library (see also Diagrid's docs), which builds MAF agents directly atop Dapr Workflows/pub-sub/state-store for durability. Pushed a follow-up commit correcting the Key findings, CF relevance, and Open questions sections accordingly. Microsoft's own first-party durable extension still uses Azure Durable Task Framework rather than Dapr — that part stands — but the picture is more nuanced than originally written.

[rkoster]

Follow-up: checked diagridio/python-ai (the Python counterpart to dotnet-ai) — it does not cover Microsoft Agent Framework. It wraps 11 other frameworks in Dapr Workflows (LangGraph, CrewAI, Google ADK, Strands, PydanticAI, OpenAI Agents, Claude Agent SDK, LangChain, Smolagents, LangChain Deep Agents, HolmesGPT) as part of Diagrid's "Catalyst" managed Dapr Workflow product, but MAF's Python binding is notably absent — only the .NET binding has a Dapr-native durability path via dotnet-ai. Pushed a follow-up commit adding this asymmetry to Key findings/CF relevance/Open questions.