fit-model: Replaced affine camera estimation with nonlinear algorithm (allows direct recovery of the pose angles)

This should make the app easier to understand, i.e. remove anything that's not necessary.

To facilitate this, added a get_texture_coordinates() method to MorphableModel

Also updated the documentation of Mesh.

- We now use the same values and formulas as in Smith & Aldrian for the detector 2D variance.
- This changes the ranges of values a bit, so set an equivalent reasonable regularisation default value

Changed the detector variance in the shape fitting from clip-coords to image-coords. This makes the regularisation behave normal again.

Also chose a slightly higher regularisation value to stay a bit closer to the mean face by default.

In the default setting, it links to OpenCV's debug libs, and that doesn't work with the STL of the RelWithDebInfo build.

Also removed a few trailing whitespaces

For path operator +=

Removed all non-header-only CMake stuff and added a custom target for the library headers so they show up in IDEs