skm_mlp_cpp: skm_mlp_cpp
In skm: Selective k-Means

solve skm with multiple runs in serial and return all w. optim

1	skm_mlp_cpp(x, k, s_must, max_it, max_at)

`x`	an m x n matrix often m < n, as a convention index rows of x with s, and cols of x with t so x(i, j) can be expressed as (s_i, t_j) equally.
`k`	number of index to be selected from x row index start from 0.
`s_must`	an index vector set should be selected before selecting other index.
`max_it`	max number of iterations can run for optimizing result. max number of iterations within a single initial run on optimal path.
`max_at`	max number of attempts or repeats on running for optimial results, max number of random initialization for finding optimial results.