Docplex utils methods to generate square matrix variables

[gcattan]

Implements #28.
The PR introduces helper methods to generate covariance matrix variables for optimization problems.
This is a pre-request to build a custom optimizer we could use to run a version of the MDM algorithm rewritten with the constraint programming paradigm.
We do this because quantum optimization algorithms in Qiskit accepts optimization problem written in docplex as possible input.

[toncho11]

I think it is OK. You can merge it.
Next step is to make a optimizer that uses these methods.

And then re-formulate:

• the Riemannian mean as an optimization problem
• the minimum distance as an optimization problem

[gcattan]

Ok, thanks for the review :)
Some of the details, such as the need for an optimizer wrapper were explained by messaging. I hope I have answered all your concerns.