An experiment at running benchmarks on multiple EC2 instances

[angelhof]

This PR includes scripts that allow running experiments on multiple ec2-instances. The infrastructure is in evaluation/multi-instance-experiment and assuming that the instances referred to in main.sh exist, we just have to call main.sh.

It now works, and can execute the eurosys one-liners experiment on two EC2 instances. The infrastructure spawns background jobs on the server and then uses a polling process to check when the experiment is done. The reason is that experiments might take a long time and if the connection drops, the experiment will stop.

What is missing:

• [DONE] Plotting results
• Refactoring the infrastructure so that it can be used for arbitrary experiments.
• [WIP] Refactor the code so creation of instances and PaSh installation is also done in a script (i.e., an installation script that: for each profile in the profiles directory creates an instance and stores its id and profile name in a line in a file, that is then read by main to execute the experiment).

Especially for the second point, I would need your @zyrolasting and @nvasilakis and that is why I added you as reviewers. Any thoughts and suggestions on how to refactor this properly would be great :)