[ENH] Add soft segment activation function for improved layer segmentation

[Leguark]

Improved Activation Functions with Soft Segmentation

This PR introduces a new soft segmentation approach for activation functions with the following changes:

• Added _soft_segment.py module with soft_segment_unbounded function that provides improved sigmoid-based segmentation
• Implemented specialized segmentation functions for lithology and fault handling
• Updated the activation interface to use the new soft segmentation by default
• Added support for both scalar and array-based sigmoid slope parameters
• Incorporated per-edge adaptive temperature calculations for better control of sigmoid transitions
• Added deprecation warning to the legacy hard sigmoid activation function
• Created comprehensive test cases to validate the new segmentation functions

The new implementation provides more precise control over the transition between geological formations and supports both scalar field values and fault handling with improved numerical stability.

[Leguark]

• [ENH] Add soft segment activation function for improved layer segmentation #11 👈 (View in Graphite)
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.