Fix httpx 'Event loop is closed' warnings during async cleanup

[DeveshParagiri]

Problem

During simulation, stderr is spammed with httpx cleanup warnings:

ERROR    Task exception was never retrieved
         future: <Task finished name='Task-177' coro=<AsyncClient.aclose() done, ...> exception=RuntimeError('Event loop is closed')>
         Traceback (most recent call last):
           File ".../httpx/_client.py", line 1985, in aclose
             await self._transport.aclose()
           ...
           File ".../asyncio/base_events.py", line 545, in _check_closed
             raise RuntimeError('Event loop is closed')
         RuntimeError: Event loop is closed

These appear hundreds of times during a simulation run.

Root Cause

The async httpx clients created by the OpenAI/Anthropic SDKs are not being properly closed before the event loop shuts down. When Python's asyncio loop closes, pending aclose() tasks fail because they try to schedule work on a closed loop.

Impact

• Functional: None - simulation completes correctly
• UX: Noisy stderr, looks like errors
• Debugging: Hard to spot real errors in the noise

Potential Fixes

1. Explicit client lifecycle: Create a single shared httpx.AsyncClient at simulation start, close it explicitly before loop shutdown

2. Context manager pattern: Wrap simulation in async context that ensures cleanup:

   async with create_llm_client() as client:
       await run_simulation(client, ...)

3. Suppress warnings: Filter these specific warnings (not ideal but pragmatic)

4. Use atexit or signal handlers: Register cleanup before loop close

Files to Investigate

• extropy/core/llm.py - where async LLM calls are made
• extropy/core/providers/ - provider implementations
• extropy/simulation/engine.py - simulation loop lifecycle

References

• https://github.com/encode/httpx/issues/2298
• RuntimeError: Event loop is closed when AsyncClient exists in multiple event_loops. encode/httpx#2959