Broadcasting `BlockedUnitRange` sometimes loses blocking information

[mtfishman]

For example (with BlockArrays v0.16.39):

julia> using BlockArrays

julia> x = blockedrange([2, 4])
2-blocked 6-element BlockedUnitRange{Vector{Int64}}:
 1
 2
 ─
 3
 4
 5
 6

julia> sin.(x)
2-blocked 6-element PseudoBlockVector{Float64}:
  0.8414709848078965 
  0.9092974268256817 
 ────────────────────
  0.1411200080598672 
 -0.7568024953079282 
 -0.9589242746631385 
 -0.27941549819892586

julia> x .+ 2
3:8

[jishnub]

This happens because of the broadcast implementations in Base that preserves ranges but not their axes. In principle, we may return

julia> BlockArray(x .+ 2, blocksizes(x,1))
2-blocked 6-element BlockVector{Int64, Vector{UnitRange{Int64}}, Tuple{BlockedUnitRange{Vector{Int64}}}}:
 3
 4
 ─
 5
 6
 7
 8

but blocksizes(x,1) will need to allocate the result. I wonder if there's a non-allocating way of obtaining the result?

Given that blockedrange necessarily sets the first element of the result to 1, perhaps this should lead to a BlockedOneTo instead of a BlockedUnitRange? In this case, we may be able to avoid the allocation, as a OneTo is its own axis.

[dlfivefifty]

A BlockedOneTo would be great to have