EDEN compliance: wf-psf modernization (Phase 1)

This milestone covers the modernization of the wf-psf package to ensure compliance with EDEN 3.1 runtime environments and compatibility with Euclid Science Ground Segment (SGS) software standards.
Objectives:

• Remove unsupported TensorFlow Addons dependencies, or move them to optional extras.
• Implement configurable optimizer selection (e.g., Adam, RectifiedAdam) to allow flexible experimentation while maintaining a fully compliant default configuration.
• Validate reproducibility and performance on simulated datasets, ensuring consistency with baseline results.
• Addition of inference package to the WaveDiff / WF-PSF framework, allowing the user to generate broadband (polychromatic) PSFs from a trained model, given a list of positions and SEDs.
• Refactors the Physical Layer Polychromatic PSF model to separate training from inference settings, enabling more flexible and modular usage across different execution contexts.