Add arange

[wolfv]

SIMDified arange can give a ~4x improvement about std::iota and for loops.

[wolfv]

The algorithm would be extremely simple:

xsimd::batch<double, 4> x(0,1,2,3);
xsimd::batch<double, 4> step(4);

for (...)
    assign (x)
    x = x + step;

needs some adjustments for step size but should all be fairly straightforward.

[serge-sans-paille]

This should go to https://github.com/xtensor-stack/xsimd-algorithm