fix: clean up garbage connections on cancellation in AsyncConnectionPool

[mbeijen]

Summary

• Fixes the AsyncConnectionPool poisoning bug reported in `httpcore.AsyncConnectionPool` poisoning due to `asyncio.CancelledError` #982: an asyncio.CancelledError (or any cancel scope timeout) during connection establishment leaves a NEW-state connection in self._connections that is neither closed nor connected, eventually exhausting max_connections and surfacing as PoolTimeout on every subsequent request.
• Adds is_connected() to (Async)ConnectionInterface, overridden on the concrete connection classes and proxy wrappers so a connection in the pre-TCP NEW state reports False.
• Teaches _assign_requests_to_connections() to recognise such garbage connections (not connected AND no live pool request references them) and silently drop them from the pool.

Ported from internal commit https://codeberg.org/httpxyz/httpcorexyz/commit/a7500e4 on the httpcorexyz fork

Reproduction (from #982)

import asyncio
import httpcore2 as httpcore

async def main():
    async with httpcore.AsyncConnectionPool(max_connections=1) as pool:
        tasks = [asyncio.create_task(pool.request("GET", f"http://localhost:8888/{i}")) for i in range(2)]
        await asyncio.sleep(0)
        for task in tasks:
            task.cancel()
        await asyncio.gather(*tasks, return_exceptions=True)

        print(pool)
        # Before: <AsyncConnectionPool [Requests: 0 active, 0 queued | Connections: 1 active, 0 idle]>
        # After:  <AsyncConnectionPool [Requests: 0 active, 0 queued | Connections: 0 active, 0 idle]>
        await pool.request("GET", "http://localhost:8888/", extensions={"timeout": {"pool": 1}})
        # Before: httpcore2.PoolTimeout
        # After:  proceeds normally (or surfaces a genuine ConnectError/ConnectTimeout)

Test plan

• scripts/check — ruff format/check + mypy clean
• scripts/test — full suite passes (1651 passed, 1 skipped) on both asyncio and trio backends
• scripts/coverage — 100% coverage maintained
• New test_connection_pool_cancellation_during_waiting_for_connection verifies the pool is empty after cancellation in the connection-wait phase, on both asyncio and trio
• Manually re-ran the reproduction from `httpcore.AsyncConnectionPool` poisoning due to `asyncio.CancelledError` #982 against the patched pool — pool stays empty and the follow-up request no longer hangs on a PoolTimeout

Closes #982.

🤖 Generated with Claude Code

[codspeed-hq]

Merging this PR will not alter performance

✅ 15 untouched benchmarks
⏩ 7 skipped benchmarks¹

---

_{Comparing mbeijen:fix-pool-cancellation-cleanup (8bfb25c) with main (c2a2b8d)}

Footnotes

1. 7 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩