Suggestion: replace Array type with numpy.typing.NDArray

[rebcabin]

So far as I can tell, the semantics are identical. I believe we ought to use a popular, existing type rather than inventing our own type, if we can.

https://numpy.org/devdocs/reference/typing.html

[certik]

The numpy.typing.NDArray semantics is different. Doesn't seem to allow specifying rank and dimensions at compile time, which is needed for ASR arrays.

If you know how to use it in our tests and preserve the information that we need, then we should definitely do it.