Loop over Iterators.rest in _foldl_impl

[jishnub]

For reasons that I don't understand, this improves performance in mapreduce in the following example:

julia> function g(A)
           for col in axes(A,2)
               mapreduce(iszero, &, view(A, UnitRange(axes(A,1)), col), init=true) || return false
           end
           return true
       end
g (generic function with 2 methods)

julia> A = zeros(2, 10000);

julia> @btime g($A);
  28.021 μs (0 allocations: 0 bytes) # nightly v"1.12.0-DEV.1571"
  12.462 μs (0 allocations: 0 bytes) # this PR

julia> A = zeros(1000,1000);

julia> @btime g($A);
  372.080 μs (0 allocations: 0 bytes) # nightly
  321.753 μs (0 allocations: 0 bytes) # this PR

It would be good to understand what the underlying issue is, as the two seem equivalent to me. Perhaps this form makes it clear that it's not, in fact, an infinite loop?

[JeffBezanson]

Huh, that's interesting (and satisfying)! The lowering of for is careful to use a branch structure that LLVM prefers, so this kind of thing is possible but I'm surprised it's still that significant. In any case the code is nicer!

[vtjnash]

In theory this change is essentially identical except that this now unrolls two iterations instead of one, so the comment above now needs to be adjusted, since it states the loop is only unrolled once