Add comprehensive documentation comments to neural network library

[dredstone1]

This PR adds detailed documentation comments to key files in the neural network library that previously lacked proper documentation. The changes enhance code readability and maintainability without modifying any functional behavior.

What was documented

Core Model Components:

• Model initialization, configuration parsing, and optimizer setup
• JSON configuration handling with error management
• Optimization algorithms including gradient descent implementation
• Console progress tracking with visual feedback

Neural Network Implementations:

• Fully Connected Networks (FNN) with forward/backward propagation
• Convolutional Neural Networks (CNN) with 2D convolution operations
• Parameter management and gradient computation
• Network visualization integration

UI Visualization System:

• Interactive button components with state management
• Real-time graph plotting for training metrics (loss/accuracy curves)
• Base panel architecture with update coordination
• Font management using singleton pattern

Documentation style

All documentation follows the existing Doxygen format used in files like tensor.cpp and activations.cpp:

/**
 * @brief Performs forward propagation through the network
 * 
 * Applies 2D convolution operation to the input using learned filters,
 * adds bias terms, and applies the activation function to produce
 * feature maps.
 * 
 * @param newInput Input tensor with spatial dimensions to convolve
 */
void forward(const global::Tensor &newInput) override;

Benefits

• Enhanced readability - Complex neural network algorithms are now clearly explained
• Better maintainability - Well-documented interfaces make the codebase easier to modify
• Developer-friendly - New contributors can understand the system architecture quickly
• Professional standards - Documentation matches industry best practices

The library now provides comprehensive guidance for understanding everything from low-level tensor operations to high-level model training workflows, making it significantly more accessible to developers working with deep learning implementations.