Log Message:
py-lmfit: updated to 1.3.3

Version 1.3.3

Fixes:

- fix loading spline models with more than five knots
- improved SplineModel to explicitly allow more knots, make it easier to \ 
evaluate and save/reload
- improvements to adding Parameters.
- support Model functions with "barestar" syntax
- fix several related problems with providing a Jacobian function, especially \ 
for consistency across solvers (including least_squares), and for pickling
- fix Step and Rectangle Models to allow a negative value for sigma, indicating \ 
a downward step

Build, Maintenance:

- update issue templates
- add test for consistent init_fit and best_fit for saved/looded SplineModel
- fix NumPy v2 DeprecationWarning
- uppdate SciPy/NumPy dependencies.
- remove numexpr dependency (again)
- drop support for Python 3.8, add Python 3.13
- asteval no longer raises NameError to Python, so we catch exceptions from \ 
asteval when creating parameters.
- avoid setting ``stderr`` to ``None`` for uncertainties calculations.

Documentation and Examples:

- add example with uncertainties in both x and y
- make sign of residual calculations in model.py consistent with documentation
- add example fitting multiple datasets using Model interface
- tweaks to 'sphinx-gallery' settings.
- update names of the documentation examples in Gallery