Convert range type in reduced_index

The `reduced_indices` and `reduced_indices0` methods sometimes assert
that the return axes type is the same as the input.
Consequently, the implementation of `reduced_index` had better
return a range of the same type as the input.

This corrects the error in
JuliaArrays/OffsetArrays.jl#92.
I'll put a workaround in OffsetArrays.jl too.

(cherry picked from commit eda4131)

Author: timholy

base/reducedim.jl

@@ -4,7 +4,7 @@
 
 # for reductions that expand 0 dims to 1
 reduced_index(i::OneTo) = OneTo(1)
-reduced_index(i::Union{Slice, IdentityUnitRange}) = first(i):first(i)
+reduced_index(i::Union{Slice, IdentityUnitRange}) = oftype(i, first(i):first(i))
 reduced_index(i::AbstractUnitRange) =
     throw(ArgumentError(
 """

test/offsetarray.jl

@@ -440,6 +440,14 @@ I = findall(!iszero, z)
 @test std(A_3_3, dims=2) == OffsetArray(reshape([3,3,3], (3,1)), A_3_3.offsets)
 @test sum(OffsetArray(fill(1,3000), -1000)) == 3000
 
+# https://github.com/JuliaArrays/OffsetArrays.jl/issues/92
+A92 = OffsetArray(reshape(1:27, 3, 3, 3), -2, -2, -2)
+B92 = view(A92, :, :, -1:0)
+@test axes(B92) == (-1:1, -1:1, 1:2)
+@test sum(B92, dims=(2,3)) == OffsetArray(reshape([51,57,63], Val(3)), -2, -2, 0)
+B92 = view(A92, :, :, Base.IdentityUnitRange(-1:0))
+@test sum(B92, dims=(2,3)) == OffsetArray(reshape([51,57,63], Val(3)), -2, -2, -2)
+
 @test norm(v) ≈ norm(parent(v))
 @test norm(A) ≈ norm(parent(A))
 @test dot(v, v) ≈ dot(v0, v0)