memusage command overhaul

memusage is not very reliable; if too much data is generated by it (this will usually be the case since allocations functions have a high call frequency) data either gets lost or corrupted ( probably since the BPF data structures used overflow); a new way that either summarizes the data in a meaningful way without saving the stacks (which I think is the problem) should be implemented. A suggestion would be to use Intel Pin, though I have not really investigated the gains.