[WIP] Fix HagerZhang bugs

[mohamed82008]

This PR closes #133, #134, and #135. There is one error I am getting in LineSearches tests which seems unrelated to the changes made. Optim tests pass. I will look into this further. Also, I need to check if other places error when the search direction is not a descent direction and try to make it robust to computational error when starting at the optimal solution. Finally, someone please comment on my use of mayterminate since I don't understand how it will be used apart from checking when the quadratic step is taken.

[mohamed82008]

I think this is good to go with the exception of some tests failing. The tests are maybe depending on the old behavior? Does that make them good tests? Any help there is appreciated. Optim tests pass, so that's a sign that I didn't totally screw up.

[pkofod]

I think that test is probably just check that behavior hasn't changed, but it seems quite likely that your changes come with changes to that exact step length. I Don't have time right now, but I'll review asap.

[anriseth]

The tests are maybe depending on the old behavior? Does that make them good tests?

Most are just to flag if we unintentionally change behaviour and are not good tests. They were added long after the implementation of the algorithms so little thought has been put into them. Feel free to update them if you believe the new behaviour is the correct one ;)

[mohamed82008]

All good now.

[codecov]

Codecov Report

Merging #136 into master will decrease coverage by 27.17%.
The diff coverage is 49.05%.

@@             Coverage Diff             @@
##           master     #136       +/-   ##
===========================================
- Coverage   69.92%   42.74%   -27.18%     
===========================================
  Files           8        8               
  Lines         389      648      +259     
===========================================
+ Hits          272      277        +5     
- Misses        117      371      +254

Impacted Files	Coverage Δ
src/hagerzhang.jl	33.9% <29.41%> (-22.72%)	⬇️
src/initialguess.jl	79.12% <58.33%> (-10.49%)	⬇️
src/static.jl	50% <0%> (-37.5%)	⬇️
src/morethuente.jl	26.84% <0%> (-32.47%)	⬇️
src/LineSearches.jl	67.85% <0%> (-32.15%)	⬇️
src/backtracking.jl	58.97% <0%> (-26.22%)	⬇️
src/strongwolfe.jl	50% <0%> (-24.29%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 1f40bf3...190ce84. Read the comment docs.

[pkofod]

Thanks! I believe code coverage is really bugged in julia these days..

[pkofod]

I've tagged a bugfix release https://github.com/notifications

[mohamed82008]

Awesome thanks for the quick merge!