Parallel matrix multiplication

Heterogeneous parallel matrix product:

1. Estimates the execution time of gsl_blas_dgemm on all processors
2. Scatters the row blocks of matrix A proportionally to the speed of processors (scatterv)
3. Broadcasts matrix B
4. Gathers matrix C (gatherv)

Implemented with the following collective communications:

• native
• Hockney_dfs_binomial_min
• Hockney_dfs_binomial_max