feat: add uncertainty quantification elastic

[Atilaac]

Introduce n_repeats to elastic_simulation to run multiple independent repeats per strain and compute uncertainty.

The function now accumulates Cij samples per repeat, returns mean and std (Cij and Cij_std) and mean/std for isotropic moduli (moduli, moduli_std).

Updated docstring/return spec and adjusted logic for normal and shear strain handling.

Tests updated/added to validate Cij index mapping, n_repeats std behavior, and related test refactors (imports and isotropic moduli checks).

Documentation and example notebook updated to document n_repeats, show printed uncertainties, and set n_repeats=3 in examples.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[Atilaac]

Here, the calculations are serial for now. In principle, this is fine since most of those are short calculations. However, I plan to run all calculations in parallel, both for deformations and for repetition, which is slightly complex. For example, we run a few independent deformations for n_repetitions, or we keep it simple and run all deformations in a single cycle for n_repetitions. This latter one, I have implemented locally. This is worth discussing, and if either is agreed upon, or if another suggestion is made, I plan to add it to another PR to keep this one focused.

[ltalirz]

Thanks @Atilaac !

Could you please make the documentation of n_repeats a bit more specific by saying what changes between repeats?

These repeats are still all for the same glass sample, right?
In your experiments, do you see significant spread between repeats?

[ltalirz]

Here, the calculations are serial for now. In principle, this is fine since most of those are short calculations. However, I plan to run all calculations in parallel, both for deformations and for repetition, which is slightly complex. For example, we run a few independent deformations for n_repetitions, or we keep it simple and run all deformations in a single cycle for n_repetitions. This latter one, I have implemented locally. This is worth discussing, and if either is agreed upon, or if another suggestion is made, I plan to add it to another PR to keep this one focused.

Ok to wait with parallelization - we first need to figure out our general approach to parallelization that works with hierarchical workflows