This is quite common in real-world code. Python 3.11 added a C API
function for this, which we use here.

The primitive works for arbtirary objects, but only type objects have
a specialized code path. Other use cases of `__name__` seem typically
less performance-sensitive.

This PR makes this micro-benchmark 2.0x faster on Python 3.13:
```
from typing import Iterator

class FooBar: pass

def foo(x: type[object], n: int) -> str:
    for a in range(n):
        s = x.__name__
    return s

def bench(n: int) -> None:
    for i in range(n):
        foo(FooBar, 1000)

from time import time
bench(50 * 1000)
t0 = time()
bench(50 * 1000)
print(time() - t0)
```