Glm class restructure

[BalzaniEdoardo]

PR Overview:

In this PR, I aim to:

1. Establish architecture principles for models within the neurostatslib package.
2. Define a NoiseModel abstract class and a concrete PoissonNoiseModel object to implement the log-likelihood and emission probability.
3. Define an abstract Solver class and various concrete subclasses to set up the optimization problems for different objective functions.
4. Provide concrete implementations of the GLM and RecurrentGLM classes that rely on solver and noise model objects.

Review Suggestions:

I recommend starting the review by reading the developer notes 02 to 05 before delving into the code. These notes provide the rationale and a description of how I structured the classes and why.

Thank you in advance for reviewing this.
Cheers,
Edoardo

[billbrod]

Looks good! Only the small changes we've already talked about.

[billbrod]

Some small typo fixes, then this is ready to go! I think the developer notes and docs will need some more work, but that's a job for a separate PR.