Remove rand hacks and use make instead

[lgoettgens]

All over AA, there are custom rand methods, that insert the RNG as a first argument, e.g.

AbstractAlgebra.jl/docs/src/rand.md

Lines 117 to 125 in 06b8630

	As mentioned above, we define a simplified random generator that saves the user
	having to create make instances.
	```julia
	rand(rng::AbstractRNG, S::PolyRing, deg_range::UnitRange{Int}, v...) =
	rand(rng, make(S, deg_range, v...))
	rand(S::PolyRing, degs, v...) = rand(Random.GLOBAL_RNG, S, degs, v...)
	```

In particular, they change rand(make(S, 1:3, -10:10)) to rand(S, 1:3, -10:10).

The RandomExtensions documentation strongly suggests to always use make in calls and not use the hack, see
https://github.com/JuliaRandom/RandomExtensions.jl/blob/f5d9723da41cb85f8c8be6ae21cb220ea3101ccc/README.md?plain=1#L35-L42

Furthermore, according to Aqua.test_ambiguity(AbstractAlgebra), these custom rand methods are part of 587 of the 622 ambiguities total!

Most of these methods have been there for at least 3 years untouched, some even longer (just been moved around between files a few times). And I think due to the above reasons, it is time for some refactoring.

[thofma]

Will this fix the problem observed in JuliaRandom/RandomExtensions.jl#13?

[lgoettgens]

Will this fix the problem observed in JuliaRandom/RandomExtensions.jl#13?

Yes, exactly that is what I meant. Thanks for pointing to the issue.

[thofma]

I am also not sure I understand the ramifications. Will we have to use rand(make(S...)) instead of rand(S,...)?

[lgoettgens]

I am also not sure I understand the ramifications. Will we have to use rand(make(S...)) instead of rand(S,...)?

As far as I have understood it (and my testing showed nothing contradicting that), this is the only visible change to the user.

[thofma]

I think writing rand(make(S,...)) instead of rand(S,...) is not acceptable. Maybe we should just ditch RandomExtensions if it makes loading everything so much slower.

[fieker]

I too think that rand(make(..)) cannot fly. We've only recently looked at the creation of random mpolys and I am afraid a normal user will not be able to create them using the current design> The proposed it worse. It's cool for experts but impossible for users

[fingolfin]

I also don't like rand(make(...)) and in light of the recent total breakage of RandomExtensions.jl (it doesn't work anymore with latest Julia nightly), I am tempted to agree that we should just ditch it.

See oscar-system/Oscar.jl#2896 for a different proposal.

[lgoettgens]

Removing these Hacks at a later stage would be a breaking change, so I think we need to address this part before 1.0 (even without a final concept and implementation of oscar-system/Oscar.jl#2896).