Update guidance for pow and divide concerning arrays with integer dtypes

[kgryte]

This PR

• supersedes Require __pow__ and __truediv__ to have floating-point inputs #221
• adds guidance for pow and divide for arrays having integer data types, as discussed in Update typing for array object dunder methods #173.
• adds a note that, when providing arrays having integer data types to pow, the result of x1_i**x2_i is implementation-defined for x2_i < 0 (e.g., 5**(-1) as the result is a fraction).
• specifies that the result of "true division" be an array having a floating-point data type.
• makes input data type guidance consistent between __pow__/pow and __truediv__/divide.

cc @asmeurer @rgommers

[kgryte]

As the changes introduced in this PR have been discussed elsewhere, will go ahead and merge. If further refinements are necessary, they can be addressed in a follow-up PR.