ParamSet$check respects defaults

[mb706]

closes #265
TODO: still missing tests

[jakob-r]

One thing came to my mind.
How would you deal with the following situation

A, default=TRUE
B, default=TRUE, requires A==FALSE
C, requires B=TRUE

Now I should not be able to set C to some value, right? You can also say that is a corner case and we should not care about such recursion and allow a false postive here.

Currently we allow defaults for "inactive" params. However I am not aware if we actually have params that depend on such params like in the example.

[mb706]

I guess we could say its an edge case and we'd like to be more permissive rather than too restrictive, but (you know me) I'll make a solution that does it right.

[jakob-r]

I just want to avoid complex code to solve such edge cases. If the requirement points on a param that is not set and has a requirement itself that is not fulfilled by the param_vals I would just stop and say infeasible. Because in such cases it might be wise that the user supplies the complete param_vals to "proof" that he knows what he is doing.