Specifically, added python bindings for:
* LandmarkMapper (core submodule)
* EdgeTopology, load_edge_topology()
* ContourLandmarks, ModelContour
* fit_shape_and_pose()
* Mesh (render submodule)

Enabling GENERATE_PYTHON_BINDINGS on travis

Previously, we would just copy the whole folders, which resulted in a quite large install folder. Now we're only copying the headers and the licence files.

Added Matlab bindings for the fitting function

For system where Matlab is not found automatically.

This mainly required conversions for glm quat and mat4x4.

This contains conversion for vector<glm::tvecX> for the vertices, colours, and texcoords, and vector<array<int, 3>> for the triangle indices.
We can now successfully return a Mesh to Matlab.

Calling it works great!
No result is returned yet from C++ to Matlab.

This works when the current working directory in matlab is eos/matlab!

Also moved the Eigen Matrix to Matlab conversion to a separate file

Fixed bug that the given num_shape_coefficients_to_fit was not forwarded to the shape-fitting function

It didn't really results in a performance improvement, but it may help.