I intentionally annotated the input types to (also) accept the overly-loose numpy.typing.ArrayLike type. Because even though that also accepts invalid input, it would avoid breaking user code like

def get_rich_quick(p: npt.ArrayLike):
    r = npf.irr(p)
    # ???
    return profit

which I'm guessing is the primary use-case of numpy_financial (or maybe that was just me).

Anyway, there's still quite a bit of room for improvement. For example, by making mypy and pyright pass is strict mode. But getting there wll take quite a bit of effort for diminishing returns.

Closes #133