When allocating a large object in gen 0, update the n_large_blocks
count.  I think this is just an accounting issue, and doesn't actually
cause a space leak, but it does result in an assertion failure when
running with sanity checking on.