TYP: append: add @deprecated overloads for python scalar coversions - #33047

I asked an AI model to try to find spots where the type changes disagree with runtime behavior. It found:

import warnings
import numpy as np

warnings.simplefilter("always", FutureWarning)
a = np.array([1], dtype=np.float32)
np.append(a, True)           # mypy: deprecated; runtime: no warning
np.append(a, np.float64(1))  # mypy: deprecated; runtime: no warning
np.append("a", 1)            # mypy: no diagnostic; runtime: FutureWarning

The True one is arguable about whether we should deprecate it (bool is a subclass of int after all). The others I think are bugs in your type changes. These three are the only class of issues I could find.

Hm good catch.
The first (True) one I think would make sense to also deprecate.
The second (float64) one is tricky because float64 subclasses float, so I'd have to hack my way around that.
The third one ("a") would be tricky to @deprecate, because np.str_ subclasses str. I could also hack around that I guess.