Provides a slow CI Slave to pinpoint test that are resource bounded

Being able to run unit tests in a slow but stable env will enable to list tests that fails because of latency (cpu/memory/disk/network) in a repeatable way.

Today we have tests that fail only when running on an overloaded slave.

This slave should:

• not be a docker
• have a slow CPU, may be 2
• a way to throttle the disk IO
• eventually add network latency (tc qdisc add dev eth0 root netem delay 4ms 1ms) but only during tests
• eventually simulate GC pause (using kill SIGSTOP/SIGCONT)