#789: Make test adjustments to backtrack nonlinear solver

[cwschilly]

Fixes #789

[cwschilly]

@eparish1 I've implemented a 2-phase approach to backtracking:

• First we try a "strict" phase with zeta = 1.0
• If that fails to converge, we revert back to the previous alpha and trialState
• Then we enter the "relaxed" phase (with the user-defined zeta)

Do you think this approach would work for the problem you described?

---

If we do stick with this approach, I think we should also add an option to enable or disable this feature (at least for testing). Right now, the unit test for the LineSearchStepTooSmall exception is failing because it can't trigger the error anymore.

To that end, could you explain briefly how input parameters to the updaters work?

I can see it that the backtracking updater takes an optional vector of scalar parameters, where right now we assume that there is only one parameter present (which defines zeta_):

BacktrackStrictlyDecreasingObjectiveUpdater(std::optional<std::vector<ScalarType> > parameter)
  {
    if (parameter)
    {
      zeta_ = parameter.value().front();
    }
  }

Should I add support for a second scalar in the vector to denote whether or not we want the two-phase approach?