Given a 2D numeric array x in which each column represents the vectorized form of an upper triangular matrix, return a 3D array y containing the concatenated triangular matrices.

• If the size of the input matrix x is MxN, then the size of the output matrix y is PxPxN, where M = sum(1:P)
• You may assume that P<=100

Example

If

x = 1  7 13
    2  8 14
    3  9 15
    4 10 16
    5 11 17
    6 12 18

then

y(:,:,1) =  1  2  4
            0  3  5
            0  0  6

y(:,:,2) =  7  8 10
            0  9 11
            0  0 12

y(:,:,3) = 13 14 16
            0 15 17
            0  0 18

NOTE: If you are wondering why this seems like a strange task, it is inspired by a genotype->phenotype mapping I am doing in a genetic algorithm.