Implement scalar functions in linalg

[imlvts]

This implements in-place implementations for scalar (element-wise) functions for linalg tensors.
The operations are applied in-place, modifying the tensor.

Example

use linalg::dense::Dense;
use linalg::scalar::ScalarTransform;

let mut a = Dense::<f32>::zeros(vec![2, 2]);
a.fill_from(&[-1.0, 4.0, 9.0, -16.0]);
a.abs();   // |x|
a.sqrt();  // √x
assert_eq!(a.data, vec![1.0, 2.0, 3.0, 4.0]);

For sparse matrix implementation, only elements which store a value are touched. This means the result is mathematically not correct for functions which don't follow f(0) = 0. If you want a mathematically correct result, the chain of applied functions needs to have zero as a fix point (f(0) = 0).