refactor: simplify convolution function signatures

[mmolari]

The convolution function had signature:

fn convolve(
    &self,
    input_grid: &Array1<f64>,
    f_values: &Array1<f64>,
    g_values: &Array1<f64>,
    output_grid: &Array1<f64>,
  ) -> Result<Array1<f64>, Report>;

But in most cases convolutions are implemented on grids with the same uniform spacing.
Moreover (except for the Riemann case) the convolution is always evaluated on the full output domain and then interpolated to fit the output grid

I think that in most cases it might be useful to simplify the signature of the convolve function to

fn convolve(&self, dx: f64, f_values: &Array1<f64>, g_values: &Array1<f64>) -> Result<Array1<f64>, Report>;

so that it takes as input simply:

• the grid spacing dx
• the two arrays with the values of the two functions on the grid
  It will then simply return the array of values of the convolution, on a grid of size |f| + |g| - 1

The restriction to particular output ranges or changes in grid sizes can then be done outside of the function. For example now the output grid determination is done here, outside of the function.

In this PR I:

• changed the signature of the convolve method, as specified above
• adapted and simplified the three specific implementations
• adapted the tests to this new signature

[mmolari]

I also changed this to use first the fine-grained dx (minimum between the two) to make the convolution more precise, and then later interpolate the result of the convolution on the coarse-grained grid (maximum dx between the two)