Tests: frontx._inverse / parameter fitting

This suite checks that inversion/fitting methods recover coherent parameters or models from data (real or synthetic) and that results are consistent with the forward solver.

What is covered

1. Parameter recovery in diffusivity models (e.g., VanGenuchten, LETd): the fit reproduces reference values within tolerances.
2. Forward/inverse consistency: when the fitted model is used in frontx.solve, it predicts the same response obtained during fitting.
3. Noise robustness: fitting remains stable with sigma ≠ 0 and respects bounds via Param(min=..., max=...).
4. Reasonable convergence: e.g., bounded number of iterations/steps.

How to run

From the repository root:

```bash pytest -q