Find a way to exercise ASLR between benchmark runs

[anp]

Respawning a process for each iteration of a microbenchmark is going to be rooough. We could probably shard the number of iterations in N buckets, where each bucket is run in a different process.