Unable to run long parallel code on more than 18 cores on Azure

The code for assessing the speedup is failing on Azure. It succeeds from 1 to 64 cores with `n = 1,000`and gave this error twice at the same spot with `n = 10,000`:

```
[1] "n_bootstrap = 10000"
[1] "num_cores = 1"
...
[1] "num_cores = 17"
[1] "Finished sampling"
[1] "num_cores = 18"
Error in mcfork() :
unable to fork, possible reason: Resource temporarily unavailable
Calls: source ... -> -> mclapply -> lapply -> FUN -> mcfork
Execution halted
Error while shutting down parallel: unable to terminate some child processes
```

Restarting the VM didn't help.

The speed-up ran successfully on 4th July 2019 and the code is exactly the same as today. I will not attempt to debug this problem.

Start with the speedup code that calls R's mclapply and reproduce the failure on Azure using n = 1,000 and n = 10,000 across increasing core counts. Inspect the mcfork resource error and child-process shutdown behavior. Done means the speedup run completes reliably beyond 18 cores, or the issue documents a confirmed platform limitation and safe workaround.

Written by the indexing model from the issue text.