BUG: QRE tracing stalling at bifurcation, incorrectly reporting output as Nash

[tturocy]

Previous joy that improvements to logit_solve handled stalling at bifurcation points appears to have been premature.

The attached game has a bifurcation point near $\lambda = 1.385313$. For a variety of parameters, the tracing cycles and stalls near this point.

As a further problem, gambit-logit is failing to check the acceptance criterion before reporting the final point as Nash; it's relying on the solver itself.

team_mp.txt