`cumsum` for `Ones` has inconsistent output type depending on element type

For example using FillArrays v1.9.3:

julia> using FillArrays

julia> cumsum(Ones{Int64}(4))
Base.OneTo(4)

julia> cumsum(Ones{Int32}(4))
4-element Vector{Int32}:
 1
 2
 3
 4

I believe the issue is this line: https://github.com/JuliaArrays/FillArrays.jl/blob/v1.9.3/src/FillArrays.jl#L578.

It may in fact be an issue with convert of Base.OneTo:

julia> convert(AbstractVector{Int64}, Base.OneTo(4))
Base.OneTo(4)

julia> convert(AbstractVector{Int32}, Base.OneTo(4))
4-element Vector{Int32}:
 1
 2
 3
 4

i.e. it seems like it should preserve that it is Base.OneTo and just change the element type. A workaround could be to use:

cumsum(x::AbstractOnesVector{II}) where II<:Integer = convert(AbstractVector{II}, oneto(II(length(x))))

Reproduce the two cumsum examples from the issue, then inspect src/FillArrays.jl around line 578 and Julia's conversion of Base.OneTo to AbstractVector. Verify the intended behavior for Int64 and Int32 and add or update coverage so the result preserves the appropriate lazy range representation.