feat(graph): support max_concurrent_jobs at the graph level

Graph agents could only inherit the app-wide job budget; the agent-level
header in graph.yaml now accepts max_concurrent_jobs alongside
model/temperature, flowing through AgentConfig::from_graph into the
run-wide supervisor. Deliberately graph-wide, not per-node: jobs outlive
the node that started them.

graph.example.yaml

@@ -36,6 +36,12 @@ top_p: null                        # Default sampling top-p for `llm` nodes
 reasoning_effort: null             # Default reasoning effort for `llm` nodes that don't override it.
                                    # Only valid when the model declares reasoning_levels.
 
+max_concurrent_jobs: 5             # Max background jobs (`job__*` tools) running at once across the
+                                   # whole graph run. Jobs live in the run-wide supervisor and outlive
+                                   # the `llm` node that started them, so the budget is graph-wide —
+                                   # there is no per-node override. Overrides the global setting;
+                                   # 0 disables background jobs for this graph agent.
+
 global_tools:                      # Tool universe an `llm` node's `tools:` whitelist draws from
   - web_search_coyote.sh
   - fetch_url_via_curl.sh