Add type NegInf representing negative infinity

[fingolfin]

Also expand the functionality of PosInf to match.

I have deliberately not yet tried to make use of this anywhere, e.g. for the degree of a zero polynomial, as that would likely be a breaking change and thus for a separate PR.

Moreover, it seems to me that this functionality actually should perhaps be moved to AbstractAlgebra... But one step after the other.

[codecov]

Codecov Report

Merging #1528 (0d042fd) into master (fd27c97) will increase coverage by 0.04%.
Report is 1 commits behind head on master.
The diff coverage is 95.45%.

@@            Coverage Diff             @@
##           master    #1528      +/-   ##
==========================================
+ Coverage   82.87%   82.92%   +0.04%     
==========================================
  Files          94       94              
  Lines       37141    37162      +21     
==========================================
+ Hits        30782    30817      +35     
+ Misses       6359     6345      -14     

Files Changed	Coverage Δ
src/HeckeMiscInteger.jl	16.21% <ø> (+0.10%)	⬆️
src/HeckeMiscInfinity.jl	95.12% <94.59%> (-4.88%)	⬇️
src/arb/arb.jl	88.30% <100.00%> (+0.57%)	⬆️
src/flint/fmpq.jl	85.74% <100.00%> (+0.14%)	⬆️
src/flint/fmpz.jl	91.22% <100.00%> (+0.18%)	⬆️

... and 5 files with indirect coverage changes

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[thofma]

Looks good. You added some signbit methods, which are not used. Could one just loop over more elements in one of the tests?

[fingolfin]

I removed Base.signbit(x::acb), as there also is no signbit for Complex (though there is a sign method with weird semantics).

Due to the funky official semantics for sign I also removed these (they are actually for an extension of ours which takes an additional argument; but I did not find a definition for them)

Base.sign(::Type{Int}, ::PosInf) = +1
Base.sign(::Type{Int}, ::NegInf) = -1

For the rest I added more tests. I also added isfinite / isinf methods for QQFieldElem (and fixed isinf(::ZZRingElem), oops).

The existing tests for fmpz, fmpq etc. seem to be rather flaky, too. E.g. things like iseven are often tested with a single input (if in doubt, always an input that leads to true as output).

[fingolfin]

@thofma OK now?

[lgoettgens]

Moreover, it seems to me that this functionality actually should perhaps be moved to AbstractAlgebra... But one step after the other.

I can try to include this in #1531 in the upcoming days.

[thofma]

I overlooked one failure in the doctests of Oscar. I mark this here as breaking.

[fingolfin]

Ahhhh, the printing changed and one doctests prints PosInf(), dang