Add complex number support to isfinite

[kgryte]

This PR

• adds complex number support to isfinite by documenting special cases. Namely, if both the real and imaginary component are finite numbers, the result should be True; otherwise, if a component is either +-infinity or NaN, the result should be False. This follows Python, Julia, and NumPy.
• updates the input and output array data types to be any numeric data type, not just real-valued data types.

Notes

• In contrast to a value which, according to convention, can be both infinite and NaN (e.g., inf + NaN j), a value cannot be both finite and NaN (e.g., 4.0 + NaN j). In other words, when evaluating whether a value is infinite, we only consider components in isolation, irrespective of whether the other component is NaN; however, for the finite test, both components are considered, such that a NaN for either component results in False, irrespective of whether the other component is finite. While perhaps not surprising, one should note that we do not have uniform rules for considering whether a complex number is infinite, finite, and/or NaN.

Reference

>>> z
(4+nanj)
>>> cmath.isfinite(z)
False

>>> import numpy as np
>>> z
(4+nanj)
>>> np.isfinite(np.array([z]))
array([ False])

julia> z
4.0 + NaN*im

julia> isfinite(z)
false

[rgommers]

LGTM. This seems consistent with existing practice, and I don't think there were any remaining concerns with that behavior. In it goes, thanks @kgryte