Bug in LangJulia generation

[jarlebring]

LangJulia generator sometimes has incorrect usage of axpby!:

julia> (graph,_)=gen_ps([4.0;3.0]);
julia> gen_code("/tmp/bla.jl",graph,lang=LangJulia())
julia> include("/tmp/bla.jl");
julia> dummy(randn(3,3))
ERROR: MethodError: no method matching axpby!(::Float64, ::UniformScaling{Bool}, ::Float64, ::Array{Float64,2})
Closest candidates are:
  axpby!(::Number, ::Union{AbstractArray{T,1}, DenseArray{T,N} where N}, ::Number, ::Union{AbstractArray{T,1}, DenseArray{T,N} where N}) where T<:Union{Complex{Float32}, Complex{Float64}, Float32, Float64} at /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.5/LinearAlgebra/src/blas.jl:566

However, if we enable explicit allocation of the I-matrix it works:

julia> gen_code("/tmp/bla.jl",graph,lang=LangJulia(true,false))

julia> include("/tmp/bla.jl");

julia> dummy(randn(3,3))
3×3 Array{Float64,2}:
  0.360612  -0.800875  0.650747
 -3.79275    3.97425   0.794266
  3.81051    0.342539  3.8338

The reason is: BLAS does not support usage of the built-in I operator. This should be fixable with if-statements in execute_operation!.