[BUG] Update Objective computation

[Iroy30]

Describe the bug
The objective should evaluate to a non-nan value.

Steps/Code to reproduce bug

from cuopt.linear_programming import DataModel, Solve, SolverSettings
from cuopt_mps_parser import ParseMps
dm = ParseMps("Bug1.mps")
print("OBJECTIVE COEFFICIENTS: ", dm.get_objective_coefficients())
sol = Solve(dm)
print("PRIMAL SOLUTION: ", sol.get_primal_solution())
print("PRIMAL OBJECTIVE: ", sol.get_primal_objective())

MPS File: ObjectiveBug.mps.zip

Output:

OBJECTIVE COEFFICIENTS:  [1. 1. 1. 1. 1. 1. 0. 0. 0. 0. 0. 0.]
Solving a problem with 6 constraints 12 variables (0 integers) and 6 nonzeros
Objective offset 0.000000 scaling_factor 1.000000
Running concurrent

Dual simplex finished in 0.00 seconds
   Iter    Primal Obj.      Dual Obj.    Gap        Primal Res.  Dual Res.   Time
      0 +0.00000000e+00 +0.00000000e+00  0.00e+00   2.00e+00     2.45e+00   0.013s
PDLP finished
Concurrent time:  0.015s
Solved with dual simplex
Status: Optimal   Objective: -nan  Iterations: 6  Time: 0.015s
PRIMAL SOLUTION:  [ -0.   1.  -0.   1.   1.   1. -inf -inf -inf -inf -inf -inf]
PRIMAL OBJECTIVE:  nan

Expected behavior
Objective value should be 4.0 instead of nan

[chris-maes]

I believe the issue is caused by dual simplex presolve eliminating empty columns (even if the variable has a infinite lower and upper bound).

A similar issue can be seen on the following MPS file

NAME          GENERIC                                        
ROWS                                                         
 N  COST                                                     
 E  EQ1                                                      
COLUMNS                                                      
    X1        COST      -1                                   
    X2        EQ1       1                                    
RHS                                                          
    RHS       EQ1       1                                    
BOUNDS                                                       
ENDATA 

which encodes the problem
minimize -x1
subject to x2 == 1
x1, x2 >= 0

This problem should be classified as unbounded but is not.

[chris-maes]

This is fixed with the merge of #287