Enrich synthetic data set with shared blocks

[ppebay]

The goal of this issue two-fold:

• create test case for full work model (including memory constraints and homing);
• further test the data generator of #506

This is an enrichment as follows: 3 equally-sized shared memory blocks, each homed to the rank with the same index, with all objects of said ranks partaking into their homed block, as follows:

The salient points of this configuration as are as follows:

1. the homed blocks will increasingly anchor the objects to their initial rank assignments, as $\delta$ will increase, hence counteracting the load-balancing pull;
2. the shared block relationships form 3 distinct object connected components: (0,1,2,3), (4,5,6,7), and (8), thus counteracting the 2 distinct object communication connected components: (0,2,3,5,6,7,8), and (1,4), depending on the relative values of $\beta$, $\gamma$ , and of the upper bound on rank memory.

[ppebay]

As per discussion with @lifflander and @nlslatt, test case is modified as follows:

in order to allow for full-cluster swaps to take place (as opposed to requiring sub-clustering when memory limit is tight).