Distmem backend does not schedule correctly

Suppose we have two nodes and three threads per node. Then we want scheduling

------------------------------
N(0) T(0)
------------------------------
N(0) T(1)
------------------------------
N(0) T(2)
------------------------------
N(1) T(0)
------------------------------
N(1) T(1)
------------------------------
N(1) T(2)
------------------------------

But what currently happens is

    indout ("SAC_WL_SCHEDULE_START( %d) = SAC_CLIP_LB0 (%s);\n",
            0, lower_bound[0]);
    indout ("SAC_WL_SCHEDULE_STOP( %d) = SAC_CLIP_UB0 (%s);\n",
            0, upper_bound[0]);

where lower_bound, upper_bound are computed by the multithreaded backend, and the CLIP macro's intersect [lower_bound, upper_bound) with [ShrayStart, ShrayEnd). So this gives actual scheduling

------------------------------
T(0)
T(0)
------------------------------
T(1)
T(1)
------------------------------
T(2)
T(2)
------------------------------

                                / \
                              /    \
                             /      \
                      Node 0 v       v Node 1
------------------------------     
T(0)
T(0)
------------------------------
T(1)
                                T(1)
                                ----------------------
                                T(2)
                                T(2)
                                ----------------------

which we do not want because it does not use all threads. Instead, we want it the other way around: first clip on the Shray level, and then pass it to the multithreaded scheduler.