Handle unstable objective function output

[jmejia8]

Improve error reporting for cases like:

julia> f(x) = rand(rand(1:3)), [0.0], [0.0]
f (generic function with 1 method)

julia> optimize(f, [-1 -1 -1;1 1 1.0])

[AlexanderNenninger]

This question is also relevant for noisy functions? Are there any heuristics that can deal with them?

[jmejia8]

No, only for cases when the user defines an objective function which returns arrays of different sizes. At the current time, implemented heuristics can solve deterministic functions. However, they can return a candidate solution.