New matrix variable functionality

[Joao-Dionisio]

What is this?

This PR will track the implementation of the matrix variable functionality. With the current changes, users can do the following:

from pyscipopt import Model, sin, log
import numpy as np

m = Model()
matrix_variable1 = m.addMatrixVar( shape=(3,3), vtype="B", name="test", obj=np.ones((3,3)) )

obj = np.asarray([ [2,1,3], [1,1,0], [1,2,3] ])
matrix_variable2 = m.addMatrixVar( shape=(3,3), vtype="C", name="", obj=obj )
# # note that depending on the argument type, what you pass to the function may be broadcasted by numpy (intentional decision on our part)
# # e.g., passing vtype="C" makes every variable have type "C", but obj is taking an iterable, and so every variable will be assigned its own value

m.addMatrixCons( log(log(matrix_variable1)) - sin(matrix_variable2) >= 32 ) # same here. This means every entry should be >= 32

m.optimize() # should be running without errors

Motivation

In some optimization sub-fields and other adjacent fields, it is sometimes standard to have a matrix view of the problem. For example, in machine learning one often gets a matrix, does some things to it, and passes it on.
It is also something that general modeling frameworks tend to support, so adding this functionality should make integrating PySCIPOpt into them easier, thus possibly increasing the visibility of PySCIPOpt.

Status

It is now possible to add matrix variables, create expression involving matrix varaibles, and create linear and general nonlinear expressions involving matrix variables. So, in a sense, most of the basic functionality is already here. What is yet missing is proper documentation, more thorough testing (especially on the nonlinear side), and a focus on performance optimization when it comes to building big models. We also want to allow addCons to accept matrix variables, as some users might find it more intuitive.

---

Please let us know your thoughts and suggestions regarding this functionality.

This is joint work by @Opt-Mucca, @mmghannam, and @Joao-Dionisio.

[codecov]

Codecov Report

Attention: Patch coverage is 42.40964% with 239 lines in your changes missing coverage. Please review.

Project coverage is 52.01%. Comparing base (1db095e) to head (f9c1caa).
Report is 62 commits behind head on master.

Files with missing lines	Patch %	Lines
src/pyscipopt/scip.pxi	36.07%	179 Missing ⚠️
src/pyscipopt/matrix.pxi	39.56%	55 Missing ⚠️
src/pyscipopt/expr.pxi	90.69%	4 Missing ⚠️
src/pyscipopt/_version.py	0.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #957      +/-   ##
==========================================
- Coverage   52.54%   52.01%   -0.54%     
==========================================
  Files          20       22       +2     
  Lines        4345     4893     +548     
==========================================
+ Hits         2283     2545     +262     
- Misses       2062     2348     +286     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.