Run some experiments concurrently

```
Some experiment types would be safe to run concurrently with each other
(notably allocation measurements). The caliper runner should be able to launch
multiple allocation benchmarks concurrently based on a configurable user
parameter, this could (at the cost of system load) decrease the latency of
finishing a single caliper run.

We may even want to allow running micro benchmarks concurrently (to fully
utilize multi-core machines), so maybe this could be configured as an
instrument option and as a global option e.g.

instrument.runtime.options.maxConcurrentMeasurements=2

This could greatly reduce the latency of performing a full measurements with
many options. The main risk is (obviously) that various experiments may
interfere with each other, it's not clear if there would be an obvious
mechanism for detecting this situation (since the runner is on the same machine
also) so we may just have to trust the user to configure things appropriately.

by default it would be nice to run multiple allocation experiments concurrently.
```

Original issue reported on code.google.com by `lu...@google.com` on 23 Aug 2013 at 4:03