[BUG] Broadcasting a vector into an `n x 1` block sparse matrix fails

For example:

julia> using BlockSparseArrays

julia> a = BlockSparseArray{Float64}(undef, [2, 2], [1]);

julia> a .= randn(4)
ERROR: BoundsError: attempt to access Tuple{Block{1, Int64}} at index [2]
Stacktrace:
  [1] getindex
    @ ./tuple.jl:33 [inlined]
  [2] (::BlockSparseArrays.var"#subblocks##0#subblocks##1"{Tuple{…}, Tuple{…}, Block{…}})(dim::Int64)
    @ BlockSparseArrays ~/.julia/packages/BlockSparseArrays/gEM83/src/BlockArraysExtensions/BlockArraysExtensions.jl:566
  [3] ntuple
    @ ./ntuple.jl:19 [inlined]
  [4] subblocks(axes::Tuple{BlockedOneTo{…}, BlockedOneTo{…}}, subaxes::Tuple{BlockedOneTo{…}, BlockedOneTo{…}}, block::Block{1, Int64})
    @ BlockSparseArrays ~/.julia/packages/BlockSparseArrays/gEM83/src/BlockArraysExtensions/BlockArraysExtensions.jl:564
  [5] (::BlockSparseArrays.var"#subblocks##2#subblocks##3"{Tuple{…}, Tuple{…}})(block::Block{1, Int64})
    @ BlockSparseArrays ~/.julia/packages/BlockSparseArrays/gEM83/src/BlockArraysExtensions/BlockArraysExtensions.jl:574
  [6] MappingRF
    @ ./reduce.jl:92 [inlined]
  [7] _foldl_impl
    @ ./reduce.jl:50 [inlined]
  [8] foldl_impl
    @ ./reduce.jl:40 [inlined]
  [9] mapfoldl_impl
    @ ./reduce.jl:36 [inlined]
 [10] _mapreduce_dim
    @ ./reducedim.jl:331 [inlined]
 [11] mapreduce
    @ ./reducedim.jl:326 [inlined]
 [12] subblocks
    @ ~/.julia/packages/BlockSparseArrays/gEM83/src/BlockArraysExtensions/BlockArraysExtensions.jl:573 [inlined]
 [13] blocked_cartesianindices
    @ ~/.julia/packages/BlockSparseArrays/gEM83/src/BlockArraysExtensions/BlockArraysExtensions.jl:580 [inlined]
 [14] #union_stored_blocked_cartesianindices##0
    @ ~/.julia/packages/BlockSparseArrays/gEM83/src/blocksparsearrayinterface/map.jl:146 [inlined]
 [15] map(f::BlockSparseArrays.var"#union_stored_blocked_cartesianindices##0#union_stored_blocked_cartesianindices##1"{…}, t::Tuple{…})
    @ Base ./tuple.jl:359
 [16] union_stored_blocked_cartesianindices(::BlockSparseMatrix{…}, ::Vararg{…})
    @ BlockSparseArrays ~/.julia/packages/BlockSparseArrays/gEM83/src/blocksparsearrayinterface/map.jl:145
 [17] (::FunctionImplementations.Implementation{…})(f::typeof(identity), a_dest::BlockSparseMatrix{…}, a_srcs::BlockMatrix{…})
    @ BlockSparseArrays ~/.julia/packages/BlockSparseArrays/gEM83/src/blocksparsearrayinterface/map.jl:89
 [18] map!
    @ ~/.julia/packages/BlockSparseArrays/gEM83/src/abstractblocksparsearray/map.jl:44 [inlined]
 [19] (::FunctionImplementations.Implementation{…})(dest::BlockSparseMatrix{…}, bc::Base.Broadcast.Broadcasted{…})
    @ BlockSparseArrays ~/.julia/packages/BlockSparseArrays/gEM83/src/blocksparsearrayinterface/broadcast.jl:81
 [20] copyto!
    @ ~/.julia/packages/BlockSparseArrays/gEM83/src/abstractblocksparsearray/broadcast.jl:54 [inlined]
 [21] materialize!
    @ ./broadcast.jl:905 [inlined]
 [22] materialize!(dest::BlockSparseMatrix{…}, bc::Base.Broadcast.Broadcasted{…})
    @ Base.Broadcast ./broadcast.jl:902
 [23] top-level scope
    @ REPL[56]:1
 [24] top-level scope
    @ REPL:1
Some type information was truncated. Use `show(err)` to see complete types.

julia> a .= randn(4, 1)
2×1-blocked 4×1 BlockSparseMatrix{Float64, Matrix{Float64}, …, …}:
 -2.23478850929105   
  0.35982400065646547
 ────────────────────
 -0.13792838702895935
 -0.0467704602085825 

See also #198, #199.

Reproduce the failing a .= randn(4) case and inspect src/BlockArraysExtensions/BlockArraysExtensions.jl around subblocks (lines 564-580), then follow the union_stored_blocked_cartesianindices and broadcast paths named in the stack trace. Done means broadcasting a vector into an n×1 block sparse matrix succeeds as the equivalent (n, 1) matrix case does, with regression coverage for the reported example.