Numpy dtype.names seen as None

```python3
import numpy as np

_DTYPE = np.dtype([("id", np.int32), ("name", np.str_, 16)])

for n in _DTYPE.names:
print(n)
```

I get on the `for` line: **Object of type "None" cannot be used as iterable value**

names is not None and it does print "id\nname" as expected.

pyright version: 1.1.411

Start by running the provided Python and NumPy reproducer with Pyright 1.1.411, then trace how dtype.names is typed and checked at the for line. Done means Pyright no longer reports that names may be None while preserving correct diagnostics for genuinely optional iterables.

Written by the indexing model from the issue text.