Move noise sampling to pulser-core

[a-corni]

Currently, each emulator has to implement its custom noise sampling. They are generally inspired from pulser-simulation. If we want to support a new noise type, we have to implement it in each emulator. This proposes to move the tools to generate samples with noise to pulser-core.

[a-corni]

For this changes to be relevant, it was convened the final object should contain:

• The noiseless samples (for plotting)
• The noise trajectory (for plotting with the noise, which is a desired new feature)
• The interaction somehow (either the register positions, or the interaction matrix directly)
• information about what type of interaction term to do (rydberg, XY, etc.)
• information about SLM mask, because the interaction matrix can change because of it

[HGSilveri]

Just a preliminary consideration on this, I'd love to discuss it further when we get the chance

[HGSilveri]

At a glance, it sounds more natural to me that this would be called HamiltonianData and given to Hamiltonian, rather than its base class. Of course, the design of the class itself would have to be adapted to this paradigm.

[a-quelle-pasqal]

I'm inclined to agree here. It's not a big change to make.