Fix power of Integer with modulus 0

[MajoranaOedipus]

When calling Integer.__pow__ with modulus 0, Mod(left, modulus) is an Integer, instead of IntegerMod_int, hence no (Mod(left, modulus) ** right).lift method or calls the wrong lift method, e.g.

pow(ZZ(-1), QQ((1, 2)), 0)  # => x, a polynomial, since it calls I.lift
pow(ZZ(3), 2, 0)  # => AttributeError: 'sage.rings.integer.Integer' object has no attribute 'lift'

I add a guard against modulus == 0 so that pow(n, ext, 0) would be the same as pow(n, ext). A related doctest is added.

Fixes #41692

📝 Checklist

• The title is concise and informative.
• The description explains in detail what this PR is about.
• I have linked a relevant issue or discussion.
• I have created tests covering the changes.
• I have updated the documentation and checked the documentation preview.

⌛ Dependencies

[github-actions]

Documentation preview for this PR (built with commit e6673c6; changes) is ready! 🎉
This preview will update shortly after each push to this PR.

[yyyyx4]

For what it's worth, Python throws a ValueError for pow(a,e,0). If you want zero to mean "no modular reduction", then a**e is the standard way to get that. I currently don't see a value in supporting this weird syntax and would therefore argue that a ValueError is the way to go.

[MajoranaOedipus]

I think raising a ValueError is also fine.

We can also raise an ZeroDivisionError, or we can call with:

if modulus == 0:
    return pow(left, right) % modulus

so that we might get consistent with (left**right)%modulus?
For example, when right or modulus is a RealNumber, the result is NaN instead.
This makes sense because for inexact calculus.
Other result types might raise their own error message.

Also there are cases when m in n % m can be a non-integer, for example when m is a RealNumber.
There was some efforts in standardizing the behaviour (#21747) although they got lost in the river of time.
m in n % m can also be a polynomial, although I doubt people would ever write pow(n, expt, m_poly) (maybe they forget to coerce n to zero-degree polynomial?).
If we like to take care of that, we can expect a TypeError in Mod(left, modulus) ** right, and fall back to (left ** right) % modulus.

To me, that Mod(n, 0) is an Integer is a surprise to me. But this is consistent with Zmod(0) == ZZ.

So ... we can do

1. raise ZeroDivisionError or ValueError directly
2. Let (left ** right) % modulus do the job (for zero modulus and maybe non-integer modulus?)
3. Make pow(left, right, 0) the same as pow(left, right)

What do you think is better?

[cxzhong]

LGTM left this to @yyyyx4