remove rmul diagonal specialization (#319)

src/generic/broadcast.jl

@@ -827,29 +827,6 @@ function copy(bc::Broadcasted{BandedStyle, <:Any, <:Any, <:NTuple{2,AbstractMatr
     _banded_broadcast(bc.f, bc.args, MemoryLayout.(typeof.(bc.args)))
 end
 
-###
-# Special cases
-###
-
-#=
-    B::BandedMatrix * D::Diagonal for DenseColumnMajor storage may be evaluated as
-    B.data .* D.diag', instead of B .* D.diag'
-=#
-function copy(M::Rmul{BandedColumns{DenseColumnMajor}, DiagonalLayout{DenseColumnMajor}})
-    B, D = M.A, M.B
-    d = diagonaldata(D)
-    Bd = bandeddata(B)
-    Bdata = if size(Bd, 1) == 1
-                # performance optimization
-                # may not be necessary in the future if the other branch
-                # is equally fast
-                reshape(vec(Bd) .* d, 1, :)
-            else
-                Bd .* d'
-            end
-    _BandedMatrix(Bdata, size(B,1), bandwidths(B)...)
-end
-
 ###
 # broadcast bandwidths
 ###