fix(runtime): use running loop for futures

[Rolly Calma (Ghraven)]

Problem

Several runtime send_message paths create futures from inside async methods with asyncio.get_event_loop().create_future().

Since these methods are already executing in a running event loop, asyncio.get_running_loop() is the more direct API and avoids relying on event-loop policy fallback behavior.

Before / after

Before, the single-threaded runtime and gRPC runtime paths asked the event-loop policy for a loop before creating response futures.

After, they create those futures from the currently running loop. The surrounding request tracking, cancellation linking, and background task behavior are unchanged.

Verification

• python -m py_compile python/packages/autogen-core/src/autogen_core/_single_threaded_agent_runtime.py python/packages/autogen-ext/src/autogen_ext/runtimes/grpc/_worker_runtime_host_servicer.py python/packages/autogen-ext/src/autogen_ext/runtimes/grpc/_worker_runtime.py
• git diff --check