Release v0.3.10

This version improves the cluster job handling by utilizing MPI based multi-node computing (a mini virtual cluster so to speak). An option nodes_per_job was added to the cluster configuration template -- when, for example, nodes_per_job: 10 is set, the number of submitted job should be 10 times less than before due to internal parallelization across nodes. This mechanism makes the large benchmark execution behavior much more friendly on clusters because this will submit much less PBS jobs to execute benchmarks.