support @ operator (__matmul__) for np.array (as alternative to using np.matrix / CyLPArray)

[sdementen]

With the introduction of the @ operator in python & numpy, it may be more natural to use standard np.array instead of np.matrix and CyLPArray to build a model.

Currently, to build CyLPExpr, we can/must use:

import numpy as np
from cylp.py.modeling.CyLPModel import CyLPModel, CyLPArray

model = CyLPModel()
T = 3
x = model.addVariable("x", T)
c = np.ones((T,))
A = np.ones((5, T,))

print(CyLPArray(c) <= x)
print(x <= c)     # no need to convert to CyLPArray as CyLPar.__le__ has priority over np.array
print(x * np.matrix(c))
print(np.matrix(c) * x)
print(np.matrix(A) * x)

(I have explicitly converted np.array to np.matrix/CyLPArray when stricly needed).

It would be nice to support the following:

import numpy as np
from cylp.py.modeling.CyLPModel import CyLPModel, CyLPArray

model = CyLPModel()
T = 3
x = model.addVariable("x", T)
c = np.ones((T,))
A = np.ones((5, T,))

print(c <= x)          # no need to wrap c into a CyLPArray
print(x <= c)
print(x @ c)            # use of @ operator for matrix multiplication instead of 
print(c @ x)
print(A @ x)

Supporting this requires to add the following to the CyLPExpr class

    __array_ufunc__ = None
    def __rmatmul__(self, other):
        v = CyLPExpr(opr="*", left=other, right=self)
        v.expr = v
        self.expr = v
        return v

    def __matmul__(self, other):
        v = CyLPExpr(opr="*", left=self, right=other)
        v.expr = v
        self.expr = v
        return v

I will prepare a PR with this change