Add @might_produce macro

[penelopeysm]

This PR introduces a macro to generate a superset of the Libtask.might_produce methods needed for a function f that might contain a produce call. For example, for a function

function f(x::Int, y::Int)
    produce(x + y)
end

instead of writing

Libtask.might_produce(::Type{<:Tuple{typeof(f),Int,Int}}) = true

you can just write

Libtask.@might_produce(f)

This works for functions with multiple methods and keyword arguments.

See #197 for the original motivation.

[github-actions]

Libtask.jl documentation for PR #198 is available at:
https://TuringLang.github.io/Libtask.jl/previews/PR198/

[penelopeysm]

CI failures

The AdvancedPS "integration test" is failing, because AdvancedPS CI itself is failing, because TuringLang/SSMProblems.jl#108

The benchmark test is failing because this is the first time it's ever been run with the new version of Turing that supports it. A lot of the code in there is not even updated for the current version of Libtask. It needs to be updated, but I'd rather not do it in this PR.

The rest are Julia 1.12.

[penelopeysm]

This line is a little bit of a sledgehammer: we're basically saying, 'any invocation of f with any positional arguments might produce'. This is not necessarily true because some methods of f might produce and some might not.

But since there isn't any real downside to marking all methods are produceable, I don't think this is a huge issue. And if someone wants to be surgical, they can still use the non-macro version.