two problems with LU for rectangular matrices

[bsxfan]

One of these problems has already been reported in the still open issue #2371, although the behaviour reported there was somewhat different from what I see now.

As far as I can see, F = lufact(randn(2,3)) is successful---it stores the L and U factors together in a 2 x 3 matrix. But there seems to be a little bit of finger trouble with extracting L:

julia> F[:L]
 ERROR: argument dimensions must match
  in promote_shape at operators.jl:157
  in + at array.jl:935
  in getindex at linalg/factorization.jl:151  

For this case, F[:U] and F[:P] work. But for F=lufact(randn(3,2)), F[:P] also fails.

det(LU) fails to check that the matrix is square. It should behave the same as det(randn(2,3)), which crashes as it should with ERROR: DimensionMismatch("det only defined for square matrices").

[ViralBShah]

Cc: @andreasnoackjensen

[andreasnoack]

I have fixed the LU issues in #3315

[ViralBShah]

This is already fixed. Closing.