Estimate minimal eigenvalue of quadratic cost hessian

[Bambade]

Add a helper function "estimate_minimal_eigen_value_of_symmetric_matrix" for estimating the minimal eigenvalue of symmetric matrix. In the sparse case it uses a power iteration algorithm, whereas in the dense case an option enables using also an exact method from EigenSolver.

This feature can be used for solving non-convex QPs, by automatically calibrating the primal proximal step size accordingly (in order to be strictly larger than this minimal eigenvalue).

The code is unit-tested in C++ and Python for both sparse and dense QPs. It is also documented with simple examples.