MonteCarloWithLocalOptimization does not pass ScoringFunction to local optimizer

[benmwebb]

Consider the following:

    lopt = IMP.core.ConjugateGradients(model)
    opt = IMP.core.MonteCarloWithLocalOptimization(lopt, 1000)
    opt.set_scoring_function(sf)
    opt.optimize(100)

The local optimizer lopt will actually use model's implicit scoring function by default. In most cases, it should probably default to using the same as opt.