[SMALL IMPROVEMENTS] When solving dual LP switch infeasible and unbounded return flags. In all equilibrium algorithms check whether the number of contacts is 0, in which case return appropriate result (typically infeasible). Improve test_static_equilibrium.

[IMPORTANT CHANGE] Take com as a 3d point rather than a 2d point (this should allow to consider cases in which gravityis not aligned with the z axis in the future).

Assume all contacts have same friction coefficient (change signature of setNewContacts method). Convert b0 (the minimum contact force generator coefficient) to e_max (the norm of the minimum force error that results in a violation of the friction cones).

Return the status of the LP solver rather than returning a boolean, so that the user can have a better understanding of what happened when the LP could not be solved.

Add new LP problems in test_data folder to use them in test_LP_problem. Change range of admissible pitch-roll angles for random contact generation from 0 to twice the friction cone angle. Clean code.

Add binary files describing some LP problems in folder test_data. Add test LP solver using binary files.

Extend code to handle arbitrary number of generators per contact. Minor improvements here and there.

[MAJOR IMPROVEMENTS] When using qpOases in DLP check for large negative objective value, which suggests likely unboundedness of the problem. Change signature of most methods of StaticEquilibrium class so that they return true if operation succeeded, false otherwise. Add tests of LP formulations against polytope projection. Add ifdef CLP_FOUND everywhere in the code so that solver CLP is now optional (still to fix the cmakelist though). Add documentation.

Add algorithms (LP and dual LP) to compute extremum equilibrium point over a line. Add relative tests (successful). Clean up old tests for computation of equilibrium robustness.

Implement LP solver using qpOases. Implement a more efficient primal formulation of the LP to compute robustness of a given CoM position (LP2 in the code). Even this formulation sometimes gives wrong results with coin-or solver, but not with qpOases. Implement first performance tests: qpOases seems faster than coin-or solver and fastest formulation seems to be the dual LP (about 0.1 ms for 32 generators). PP algorithm is instead very slow, much slower than in Steve's code, but it's not clear why.