Proposal for meta ini syntax: "local variables"

The discussion on !25 (merged) made me think about the semantics of defining a system test through a meta ini file and whether our syntax is mighty enough. So, here is my proposal.

I would like to enforce the following semantics on a meta ini file:

• A meta ini with the static group whitelisted should describe the set of static variations.
• A meta ini with the static group blacklisted should describe the set of dynamic variations.

http://conan2.iwr.uni-heidelberg.de/git/dominic/dune-pdelab-systemtesting/blob/master/dune/pdelab-systemtesting/poisson/poisson.mini for example is okay with 1) (because we have been using this extensively) , but it violates 2), because it defines variables in the "dynamic section", which refer to static variations.

As it might be cumbersome to fully separate the two, I want to propose instead, that we define any key starting with a (single) underscore to be a local key of the meta ini file. Properties of such local key:

• It will always be deleted on the filtering
• As a side effect of above, it will never influence the amount of configurations
• it is documented that such key is not usable from the actual executable.