Michael Baudin

@josephmure : I took into account your review comments : thank you for this. Please review the changes and let me know.

@adutfoy > It is not necessarily equal to the normal distribution. In [^rawlings2001] page 3: "_The random errors $\epsilon_i$ have zero mean and are assumed to have common variance $\sigma^2$...

@jschueller : I think that this one is ready for merging.

> Can you address my remaining comments please? Ok: please let me have a look.

> These comments are: [#2451 (comment)](https://github.com/openturns/openturns/pull/2451#discussion_r1549659570) and [#2451 (comment)](https://github.com/openturns/openturns/pull/2451#discussion_r1549321628) @josephmure : I fixed your comments: thank your for pointing these inconsistencies. Can you tell if that corresponds to your requirements?

@jschueller I think that this PR is ready for merging.

A number of ideas are elaborated. - A script can read the code and check that each `ResourceMap` key of a given class is documented. If the number of undocumented...

The fact that a `false` boolean is used instead of a `float` is not a bug according to you? I guess that an example of this class would be welcome...

The penalized least squares problem is: $$ b = \textrm{argmin}_{b} \left(y - b^T \Psi\right)^T W \left(y - b^T \Psi\right) + \lambda b^T Q b $$ où $\lambda \geq 0$, $Q$...

There is no example of the `PenalizedLeastSquaresAlgorithm` class.