[WIP] Add support for exact newton polytope #2

[blegat]

Closes #2

• [ ] Multipartite
• [x] Exact Newton Polytope
• [ ] Tests

[blegat]

Note: motzkin polynomial is a nice test example

[chriscoey]

This would be a great feature showing off different Julia packages/interfaces working together. Also useful. For most polynomial problems that would be feasible to solve, the size wouldn't be too large for a good polyhedral computation package like CDDLib to handle, right?

[blegat]

Note that in YALMIP, this is done by default but it does not precompute the H-representation but rather get separating hyperplanes lazily by solving LP's in the V-representation. See section III.A of Pre- and Post-Processing Sum-of-Squares Programs in Practice.

We cannot really do this by default since it would require having an LP solver as a dependency but it would be nice to also have this feature. In the paper mentioned above the authors mention that there are sometimes issues with the H-representation methods.