Add A/A testing

[luabud]

A/A testing is used to validate metrics for A/B testing and/or the sampling logic itself.
Considering we have X experiments with Y% users in each group, we should take the remaining Z=100%-(X*Y)% (i.e. people who are not in any of the experiment groups) and sample them into control and treatment groups, with Y% users in each one.

[luabud]

Just to be clear, we want to separate users who are not in any A/B experiments between experiment and control groups so we can see if our sampling method is being fair. We want to make sure that there's no significant change in the behavior in each group

/cc @qubitron

[DonJayamanne]

Size - 10%

[luabud]

Validated.