I am trying to extract the `Q`

matrix from a QR decomposition without allocating. The following hack doesn’t work:

```
julia> using LinearAlgebra
julia> A = rand(3,2);
julia> qrf = qrfact(A);
julia> qrf.Q * Matrix{Float64}(I,2,2)
3×2 Array{Float64,2}:
-0.554696 0.791229
-0.64075 -0.603576
-0.530803 -0.0982461
julia> Q = zeros(3,2);
julia> A_mul_B!(Q, qrf.Q, Matrix{Float64}(I,2,2))
┌ Warning: `A_mul_B!(C::AbstractMatrix, A::AbstractVecOrMat, B::AbstractVecOrMat)` is deprecated, use `mul!(C, A, B)` instead.
│ caller = top-level scope
└ @ Core :0
ERROR: DimensionMismatch("matrix A has dimensions (3,3), matrix B has dimensions (2,2)")
Stacktrace:
[1] _generic_matmatmul!(::Array{Float64,2}, ::Char, ::Char, ::LinearAlgebra.QRCompactWYQ{Float64,Array{Float64,2}}, ::Array{Float64,2}) at C:\cygwin\home\Administrator\buildbot\worker\package_win64\build\usr\share\julia\site\v0.7\LinearAlgebra\src\matmul.jl:518
[2] generic_matmatmul!(::Array{Float64,2}, ::Char, ::Char, ::LinearAlgebra.QRCompactWYQ{Float64,Array{Float64,2}}, ::Array{Float64,2}) at C:\cygwin\home\Administrator\buildbot\worker\package_win64\build\usr\share\julia\site\v0.7\LinearAlgebra\src\matmul.jl:509
[3] mul! at C:\cygwin\home\Administrator\buildbot\worker\package_win64\build\usr\share\julia\site\v0.7\LinearAlgebra\src\matmul.jl:156 [inlined]
[4] A_mul_B!(::Array{Float64,2}, ::LinearAlgebra.QRCompactWYQ{Float64,Array{Float64,2}}, ::Array{Float64,2}) at .\deprecated.jl:57
[5] top-level scope
```

Any idea how to do this in v0.7 and preferably v0.6?