Don't print success if a stopping criterion is non-finite.

[pkofod]

Fixes #948

The new behavior is that Optim.converged returns false if any x or f change measure is nonfinite and if the gradient norm is nonfinite.

[codecov]

Codecov Report

Merging #997 (8947d71) into master (d5cb5da) will increase coverage by 0.11%.
The diff coverage is 83.33%.

❗ Current head 8947d71 differs from pull request most recent head a35518d. Consider uploading reports for the commit a35518d to get more accurate results

@@            Coverage Diff             @@
##           master     #997      +/-   ##
==========================================
+ Coverage   85.34%   85.45%   +0.11%     
==========================================
  Files          42       42              
  Lines        3187     3198      +11     
==========================================
+ Hits         2720     2733      +13     
+ Misses        467      465       -2     

Impacted Files	Coverage Δ
src/multivariate/optimize/optimize.jl	90.62% <ø> (ø)
src/types.jl	84.46% <62.50%> (+0.30%)	⬆️
src/api.jl	72.89% <100.00%> (+1.89%)	⬆️
src/utilities/assess_convergence.jl	100.00% <100.00%> (ø)
src/multivariate/solvers/constrained/samin.jl	78.57% <0.00%> (+1.19%)	⬆️

Help us with your feedback. Take ten seconds to tell us how you rate us.