Not sure yet whether it's a good design decision!

Also if we keep it this way, have to add proper getters, make member variables private, etc.

Using the new renderer stuff.

It is modeled after the vertex shader / fragment shader approach of modern OpenGL.
It also includes a texture extraction fragment shader, which hopefully will replace the current affine texture extraction.

This is still very much experimental, containing lots of todo's. For example it contains lots of if-hacks to select the correct fragment shader (for rendering or texture extraction).

Consistent 'REQUIRED' usage

Tabs to spaces in main CMakeLists file

Using 2 spaces for CMake code for now

Adjusted all includes, files and bindings.

In contrast to the MorphableModel member function draw_sample, this method takes blendshapes coefficients too and returns a full assembled mesh.

If 'REQUIRED' used inside 'find_package' there is no need to check '*_FOUND'
variable. And vice versa - if package is not optional then 'REQUIRED' can be
used instead of 'if(NOT *_FOUND) message(FATAL_ERROR)' block.

Apparently the dtype comparison compared pointer addresses, so it must've worked just by chance. Scary!

Also updated documentation

We copied this from the pybind11/eigen.h header. If we need it, we can always re-add it.

Let's see if there's ever a "modern CMake" solution for this without a custom target.
target_sources(eos INTERFACE ${HEADERS}) doesn't seem to be a proper solution.

Also swapped order of project() declaration and cmake_minimum_required()

30 is most likely a more reasonable default lambda for the new iterative shape-fitting - need to get some more experience with it

pybind11 deprecated .check().