fix: handle missing event loop in close() during interpreter shutdown

[haosenwang1018]

Closes #135

When the atexit handler calls close(), the default event loop policy may already be torn down during interpreter shutdown, causing asyncio.run() to raise RuntimeError: There is no current event loop in thread 'MainThread'.

This fix adds a fallback: if asyncio.run() fails, we manually create a new event loop with asyncio.new_event_loop() and run the finalization there. This ensures storages are properly finalized regardless of when the cleanup runs.

[LarFii]

Thanks for the fix.

I reviewed the change and found two potential issues:

1. Blocking: RuntimeError handling is too broad in the fallback path
   In raganything/raganything.py, the new except RuntimeError: around asyncio.run(self.finalize_storages()) may also catch RuntimeError raised inside finalize_storages() itself (not only event-loop-policy teardown errors).
   That can:

   • mask the original failure reason,
   • trigger a second finalize attempt unexpectedly,
   • make debugging harder.

2. Non-blocking: manual loop should be set as current loop during fallback
   In the new_event_loop() branch, it is safer to call asyncio.set_event_loop(loop) before run_until_complete(...), then clear it with asyncio.set_event_loop(None) in finally.
   Some downstream code may still call asyncio.get_event_loop() during finalize; without setting the current loop, fallback can still fail in those cases.