Agent Workflow Dependencies

You're building an AI agent orchestration system. Agents have dependencies — an agent cannot run until all its dependencies complete.

Part 1: get_execution_order(agents, dependencies) — Return a valid execution order using topological sort. Return None if there's a cycle.

Part 2: find_critical_path(agents, durations, dependencies) — Return the minimum total time to complete all agents (longest path in the DAG).

Example:

agents = ["data_fetch", "preprocess", "model_run", "postprocess"]
dependencies = [
    ("data_fetch", "preprocess"),
    ("preprocess", "model_run"),
    ("model_run",  "postprocess"),
]
get_execution_order(...) → ["data_fetch", "preprocess", "model_run", "postprocess"]