Cache sniffio availability instead of importing it on every call

[mbeijen]

current_async_library() re-imported sniffio on every invocation, falling back to "asyncio" on ImportError. Since anyio>=4.11 no longer depends on sniffio, anyone using httpcore2[asyncio] without trio installed pays a failed-import cost on every lock/event/semaphore setup, and Python doesn't cache failed imports.

Now sniffio is imported once at module load, like trio and anyio already are, and current_async_library() just checks the cached reference. Only availability is cached, not the detected backend.

Reported in encode/httpcore discussion #1102, with the fix and test strategy suggested in a follow-up comment:
encode/httpcore#1102 (comment)

Summary

improve import time when modern anyio =->4.11 is installed and sniffio isn't

Checklist

• I understand that this PR may be closed in case there was no previous discussion. (This doesn't apply to typos!)
• I've added a test for each change that was introduced, and I tried as much as possible to make a single atomic change.

[codspeed-hq]

Merging this PR will not alter performance

✅ 17 untouched benchmarks
⏩ 7 skipped benchmarks¹

---

_{Comparing mbeijen:fix/sniffio-hot-path-import (b3f847f) with main (d03f1ec)}

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. ↩

[cubic-dev-ai]

No issues found across 2 files

_{Re-trigger cubic}