LOL, another intermittent infinite loop

[erikerlandson]

PR #6 fixed a bunch of infinite looping cases but I still managed to trigger some. It turns out that the delta-x vector can go arbitrarily long without ever converging in magnitude. I suspect these behaviors are specific to the code path for equality constraints, where the dual solution space allows for more unintuitive behaviors. However I'm installing convergence checks for both code paths.

Checking for converging objective function seems to be fixing this, but I am going to add some more brute-force randomized testing to try and smoke these out. If it's the solution I think it subsumes the delta-x check, and if so I am going to remove that.