New SCIP backend implemented!

[kchanqvq]

I implemented an SCIP backend here. Feel free to advocate here in the main repo if you want! And suggest anything :)

I did reinvent premature-solution-error. I think it's good to move this definition into the main repo and make both GLPK and SCIP backend inherit from it. Let's coordinate this if you want :)

[neil-lindquist]

That's great to hear! I just added a reference to the main README.

I think that it makes sense to include a premature-solution-error in the main repo since that would increase the seamlessness of transitioning from one backend to another. Do you have thoughts on the specifics of that?

And sorry for the delay in responding to all your work. I got busy with some other things.

[kchanqvq]

I suggest

(define-condition premature-solution-error (solver-error)
  ())

in linear-programming, and let linear-programming-scip:premature-solution-error, linear-programming-glpk:premature-solution-error all inherits from this. I think backend implementers are expected to provide CONTINUE restart for this, but I don't know a way to enforce this. Say it in a comment or docstring?

I can open PRs to both linear-programming and linear-programming-glpk so you can merge in one go. Or is it possible to give me commit permission?

And sorry for the delay in responding to all your work. I got busy with some other things.

No worries, me as well! You just remind me for even larger, larger problem the alist loop in sum-linar-expressions become bottleneck as well, I'll put together a PR soon...

[neil-lindquist]

Yeah, that looks good. Maybe a doc string to the effect of (:documentation "Base type to indicates the solver exited before finding the optimal solution. A continue restart should be provided to continue with the computed sub-optimal solution.")