Intended behavior of bracket powers

Nobody has claimed this yet.

Let's focus on the characteristic-p setting, so that there is a mathematically meaningful notion of I^[p^e] for an ideal I. I would have thought that I^[p^e] is the ideal generated by p^e-th powers of the generators of I, but that is not what M2 has implemented. Is the following behavior intended?
```
S = (GF 4)[x]
I = ideal(x+a)
I^[2] -- x^2 + a, rather than x^2 + a^2 = x^2 + a + 1
radical I^[2] == radical ideal I -- fails!
```
There is a mathematically meaningful operation behind this behavior of I^[2]: it is the expansion of I along the relative Frobenius of GF 4. But I think most people in the characteristic p world would take I^[2] to mean the expansion along the *absolute* Frobenius, i.e., take p-th powers of generators (and the resulting ideal thus has the same radical as I). The existing behavior means that the usual characteristic p calculations like I^[p^e]: I occurring in Fedder's criterion give the "wrong answer" when working over GF p^e.

So, what are bracket powers intended to do?

No source file or test is named. Start by reproducing the supplied GF 4 example with bracket powers, radicals, and the Fedder-criterion expression; done means establishing and recording whether bracket powers should use relative or absolute Frobenius, and resolving the resulting behavior.

Written by the indexing model from the issue text.