Processing of {min/mean/max}.controls arguments

[benthestatistician]

findSubproblems() drops at least some empty or infeasible subproblems. When it's invoked in fullmatch.matrix(), in a situation where the user has provided a vector min.controls argument:

• [x] we need to trim the min.controls vector to match the length of the subproblem list. (And similarly for mean.controls, max.controls, omit.fraction.)

• [ ] We also need to ensure alignment of these arguments with subproblems. (I think; perhaps it's already being done in some cases. But I couldn't see where.)

• [ ] We need unit tests of this input-checking, for future proofing.

[benthestatistician]

314eee7 addresses the first part and at least some of the second. Remains to do testing.