Initial implementation of finite-difference gradient approximations for likelihood

[alexnick83]

Implements finite-difference 5-point stencil for computing the gradient and hessian of the likelihood. Also introduces small API changes, such as the following:

• The Likelihood class now has two new methods, finite_difference_gradient_likelihood and finite_difference_hessian_likelihood, that implement the approximations.
• Likelihood configuration now has a method attribute that takes values in ["exact", "finite-difference"]. The attribute can be used to switch between the analytical formula for gradients and the finite-difference approximation. Default value is "exact".
• The method attribute is used in the new methods gradient_likelihood and hessian_likelihood that call evaluate_<x>_likelihood or finite_difference_<x>_likelihood (x is gradient or hessian) accordingly.
• To facilitate the computation of the finite-difference approximation, the evaluate_likelihood method now returns the full vector instead of the sum.
• The Likelihood class now has an evaluate_sum_likelihood method that calls evaluate_likelihood and sums up the vector entries.

[Copilot]

Pull Request Overview

This PR implements finite-difference gradient and Hessian approximations for likelihood functions as an alternative to exact analytical computations. It introduces a new method configuration option to switch between "exact" and "finite-difference" methods.

Key changes:

• Added finite-difference gradient and Hessian computation methods using 5-point stencil
• Modified likelihood API to return vectors instead of scalars from evaluate_likelihood
• Updated method calls throughout the codebase to use new API methods

Reviewed Changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
src/dalia/core/likelihood.py	Core implementation of finite-difference methods and new API structure
src/dalia/configs/likelihood_config.py	Added method configuration option
src/dalia/likelihoods/*.py	Updated evaluate_likelihood to return NDArray instead of float
src/dalia/core/model.py	Updated to use new likelihood API methods
src/dalia/models/coregional_model.py	Updated to use new likelihood API methods
src/dalia/submodels/brainiac.py	Added TODO comment about likelihood specialization

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[vincent-maillou]

Should close #22