Fixes #2258

As mentioned in #2258, Eigen v3.4 supports slicing for dense matrices making many of the igl::slice overloads obsolete. I've removed all of their uses within libigl internally and in the tests and tutorial so that now dense slices show up in the static library. They're still available for now (e.g., via header only).

This will break code relying on these in the static library, but the fixes are straight forward (and result in cleaner + faster code! thanks, eigen!).

I've also added a igl::find overload to make using binary masks easy.

There's a nasty gotcha in Eigen's slicing, I'll document here:

Eigen::MatrixXd A(4,2);
A<< 1,2,
  3,4,
  5,6,
  7,8;
Eigen::Array<bool,Eigen::Dynamic,1> M(4);
M<<0,1,0,1;
// Correctly select rows using mask M
Eigen::MatrixXd B = A(igl::find(M),Eigen::all);
// Unfortunately, this also compiles:
Eigen::MatrixXd C = A(M,Eigen::all);

std::cout<<igl::matlab_format(B,"B")<<std::endl;
std::cout<<igl::matlab_format(C,"C")<<std::endl;

this results in

B = [
  3 4
  7 8
];
C = [
  1 2
  3 4
  1 2
  3 4
];

We can see that the true and false values of the Eigen::Array<bool,…> M have been interpreted as indices 0 and 1 🙃

Remembering to useigl::find is a fine solution as long as ... well ... you remember. I wish Eigen would throw a warning if you use bool scalar types as a slicing index.