CC hangs if the number of batches to send is too large

[Victor-C-Zhang]

The issue

At the beginning of a CC, we need to flush the buffers and update supernodes. It seems like this is hanging if there are not enough updates buffered. There's probably a bug in the code to dispatch a message to the MPI worker. We might need a force_send option for this.

How to replicate

• Run distributed tests with WorkerCluster::num_batches = 512
• On dist_query branch run speed experiment -np 4 ./speed_expr 50000 50012 [output file]