Add info on garbage collection of async clients

[damccorm]

• I understand that this repository is auto-generated and my pull request may not be merged

Changes being requested

When using multiple AsyncOpenAI clients to send records to a local vLLM server, we ran into an issue with garbage collection after switching from the sync client to the async client. We started seeing a bunch of errors like:

Traceback (most recent call last):\n  File \"/usr/local/lib/python3.9/dist-packages/httpx/_client.py\", line 1985, in aclose\n    await self._transport.aclose()\n  File \"/usr/local/lib/python3.9/dist-packages/httpx/_transports/default.py\", line 406, in aclose\n    await self._pool.aclose()\n  File \"/usr/local/lib/python3.9/dist-packages/httpcore/_async/connection_pool.py\", line 353, in aclose\n    await self._close_connections(closing_connections)\n  File \"/usr/local/lib/python3.9/dist-packages/httpcore/_async/connection_pool.py\", line 345, in _close_connections\n    await connection.aclose()\n  File \"/usr/local/lib/python3.9/dist-packages/httpcore/_async/connection.py\", line 173, in aclose\n    await self._connection.aclose()\n  File \"/usr/local/lib/python3.9/dist-packages/httpcore/_async/http11.py\", line 258, in aclose\n    await self._network_stream.aclose()\n  File \"/usr/local/lib/python3.9/dist-packages/httpcore/_backends/anyio.py\", line 53, in aclose\n    await self._stream.aclose()\n  File \"/usr/local/lib/python3.9/dist-packages/anyio/_backends/_asyncio.py\", line 1314, in aclose\n    self._transport.close()\n  File \"/usr/lib/python3.9/asyncio/selector_events.py\", line 698, in close\n    self._loop.call_soon(self._call_connection_lost, None)\n  File \"/usr/lib/python3.9/asyncio/base_events.py\", line 751, in call_soon\n    self._check_closed()\n  File \"/usr/lib/python3.9/asyncio/base_events.py\", line 515, in _check_closed\n    raise RuntimeError('Event loop is closed')\nRuntimeError: Event loop is closed\n

which indicated that the event loop was being closed before we could close the client. Concurrently, we saw degraded performance. When we switched to with blocks for resource management, the problem went away (PR - https://github.com/apache/beam/pull/35053/files). I seems like relying on normal garbage collection calling https://github.com/openai/openai-python/blob/f588695f77aad9279a355f5f483d8debf92b46ed/src/openai/_base_client.py#L1318C21-L1318C37 can cause a meaningful performance issue because the event loop can be destroyed before the client cleanup finishes.

Its not clear to me if the slowness was caused by:

1. Overhead associated with trying to call a closed event loop
2. Actually failing to close the connections

Regardless, it would be nice to doc this since switching clients did not work without some minor changes.