py3: fix various issues in cplex backend

[dcoudert]

Apart from the problems reported in #27773, several doctests are failing in Python 3 in src/sage/numerical/backends/cplex_backend.pyx due to:

• bytes versus str
• passing iterator (zip) instead of list in add_linear_constraint

and some issues are already here with Python 2:

• int(CPX_INFBOUND) give 0 while CPX_INFBOUND == 1E+20. Hence, some doctests are failing. Actually, the int(...) is not needed
• with a integer variable, get_variable_value return 2.0 in py2 and 2 in py3 as round has a different behavior. We unify the behavior using int(...).

[dcoudert]

New commits:

​61bf0bb

trac #27774: fix various issues in cplex backend

[dcoudert]

similar issues may also occur with gurobi...

[tscrim]

While I am not strictly opposed to these sorts of changes:

-            sage: p.add_linear_constraint( zip(range(5), range(5)), 1.0, 1.0, name='foo') # optional - CPLEX
+            sage: p.add_linear_constraint([(i, i) for i in range(5)], 1.0, 1.0, name='foo') # optional - CPLEX

I think it would be better to have add_linear_constraint handle these types of inputs.

[git]

Branch pushed to git repo; I updated commit sha1. New commits:

​2667918	trac #27774: Merged with 8.8.beta5
​c408bc0	trac #27774: make add_linear_constraint accept iterator

[dcoudert]

I did a change in all backends in which coefficients was iterated several times. May be there is a smarter way to do it, but at least this one is working.

To show that it's working, I ensured that in all the modified files, the doctest is with zip(...) and not list(zip(...)).

[tscrim]

Thank you. I think this is a much better solution (and now the behavior also matches the docstring).