Evaluates a Databricks Supervisor Agent that routes to sub-agents (Knowledge Assistants, Genie, agent endpoints, MCP, UC functions); emits route, delegation, synthesis, loop, and cost `mlflow.genai.scorer` callables plus a 3-tier per-route release gate. Use when traces have `supervisor_route` spans over sub-agent spans and a route x scenario sample exists. Do NOT use when the agent is single-agent (use `evaluating-custom-agents`) or the issue is tool runtime (use `evaluating-tool-calling`).
