TYP: use DTypeLike alias from numpy.typing

[simonjayhawkins]

No description provided.

[jbrockmendel]

do we need to worry about strings like "categorical" that can be interpreted as pandas dtypes but not np.dtype?

[Dr-Irv]

@simonjayhawkins I hope you an approve and merge in #40421 before this one gets merged, because many of the changes there are with respect to using NpDtype in the right places with respect to the to_numpy() calls in ExtensionArray classes and subclasses.

[simonjayhawkins]

@simonjayhawkins I hope you an approve and merge in #40421 before this one gets merged, because many of the changes there are with respect to using NpDtype in the right places with respect to the to_numpy() calls in ExtensionArray classes and subclasses.

I recall suggesting doing this as a precursor. some changes in #40421 would become unnecessary as the mypy errors are false postives.

[simonjayhawkins]

closing to clear queue. will hopefully circle back to typing next week.

[simonjayhawkins]

@Dr-Irv @jbrockmendel I probably won't have time to revisit this for a while (probably after 1.3.1 to allow time to follow regressions from 1.3) so feel free to take this on.

[simonjayhawkins]

I could be wrong, but I think this is necessary precursor to the EA typing work.

[Dr-Irv]

I could be wrong, but I think this is necessary precursor to the EA typing work.

I think you could do it in either order. If we can merge in the things I have been working on since late January, incorporating the changes in this PR will be straightforward. On the other hand, if we merge in this PR, then I probably have to make changes in my 4 open PRs, which would be a PITA.