What do optimizers do if a call to CostFunction::cost(...) or Gradient::gradient(...) return an error?

[bcolloran]

Hey really enjoying argmin, thanks for this crate!

As mentioned in the post title: I'm curious what happens CostFunction::cost(...) or Gradient::gradient(...) return an error. Didn't spot it in the trait docs, so thought it might be useful ask here.

In particular, I'm wondering whether solvers will generally recover gracefully, or whether the optimization will halt as soon as one hits an error value.

Thanks!