fix(transport): catch all TimeoutError variants in process termination escalation loop

[Echolonius]

🐛 Fix: Catch All TimeoutError Variants in Process Close Escalation Loop

💡 Overview & Root Cause Analysis

In src/claude_agent_sdk/_internal/transport/subprocess_cli.py, the close() method runs inside a shielded cancel scope and uses anyio.fail_after(5) to enforce a graceful shutdown timeout, escalating from wait() to terminate() and finally kill().

However, the escalation loop previously caught only built-in TimeoutError explicitly (except TimeoutError:). Depending on the AnyIO version or backend context (asyncio vs trio), anyio.fail_after can raise anyio.TimeoutError. Catching (TimeoutError, getattr(anyio, "TimeoutError", TimeoutError)) guarantees that the process termination escalation loop (terminate() -> kill()) executes reliably across all Python runtime environments.

🛠️ Solution

Updated the except blocks in SubprocessCLITransport.close() to handle (TimeoutError, getattr(anyio, "TimeoutError", TimeoutError)).

✅ Verification

• All 146 pytest tests passing cleanly in 2.89s (uv run --with pytest --with pytest-asyncio --with trio pytest tests/test_transport.py).
• GPG/SSH signed commit.