REF: standardize astype in EA subclasses

[jbrockmendel]

No description provided.

[simonjayhawkins]

Thanks @jbrockmendel, by changing the accepted types for astype we are changing the api for 3rd party EAs. probably needs a release note. (subclassed EA need to accept the same types as the base class) cc @Dr-Irv if any comments

[Dr-Irv]

Thanks @jbrockmendel, by changing the accepted types for astype we are changing the api for 3rd party EAs. probably needs a release note. (subclassed EA need to accept the same types as the base class) cc @Dr-Irv if any comments

IMHO, I think you ought to first accept/merge in #41251 where I created overloads for astype in various subclasses to separate out the cases of using a base type versus a numpy dtype, and then merge in this one.

I think there is a separate issue of how we should handle astype(extension_dtype) for crossing between extension dtypes and regular types, and vice versa, which I think this PR is trying to address.

[jbrockmendel]

ill just revert the type annotation

[jbrockmendel]

Mothballing since i cant replicate the failures locally.