Add solver dependent callbacks

[rafabench]

Close #58
Added MOI heuristics and user cuts callbacks
Also added a callback to print log in Windows

[codecov-io]

Codecov Report

Merging #85 into master will increase coverage by 2.00%.
The diff coverage is 81.01%.

@@            Coverage Diff             @@
##           master      #85      +/-   ##
==========================================
+ Coverage   58.77%   60.77%   +2.00%     
==========================================
  Files           7        9       +2     
  Lines        2559     2761     +202     
==========================================
+ Hits         1504     1678     +174     
- Misses       1055     1083      +28     

Impacted Files	Coverage Δ
src/Xpress.jl	100.00% <ø> (ø)
src/xprs_callbacks.jl	52.27% <52.27%> (ø)
src/MOI/MOI_callbacks.jl	86.72% <86.72%> (ø)
src/MOI/MOI_wrapper.jl	86.73% <89.13%> (+0.11%)	⬆️
src/api.jl	26.13% <100.00%> (+0.94%)	⬆️
src/helper.jl	42.71% <100.00%> (+1.71%)	⬆️
... and 1 more

---

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 c225699...8c79274. Read the comment docs.

[frapac]

Thanks for adding this to Xpress.jl!
In my experience, we have to be careful when using lazy callbacks with Xpress. Most of the time I have to deactivate the presolve to avoid annoying error.
E.g., testing the new callbacks you just added, I get

?405 Error: Invalid column number passed to XPRSstorecuts.
Column number 0 is invalid
ERROR: LoadError: XpressError(32): Subroutine not completed successfully, possibly due to invalid argument. Unable to call `Xpress.storecuts`:

405 Error: Invalid column number passed to XPRSstorecuts.
Column number 0 is invalid

as the presolve removes all the columns on this simple model. The problem here is that we do not know on the MOI side which variables is using Xpress after the presolve.

[joaquimg]

thanks @frapac !
This error is raised in some test already here?
Or is it something you built?

[frapac]

In this PR, the tests are carefully written and set explicitly the right set of options. I am wondering what is the correct way to signal that to the end user

• by setting explicitly PRESOLVE to 0 in the MOI wrapper
• or by adding more documentation about the callbacks in the README (a warning issued in the MOI wrapper could also do the job)

This error was raised by something I built (basically a sparse logistic regression with lazy callbacks implemented on top of Xpress). Unfortunately, I still have some trouble to make them work as expected... If you are interested, you could find the source here

[joaquimg]

some tweaks