docs: Add comprehensive PhysicsNeMo code structure documentation

[ram-cherukuri]

This PR adds a comprehensive code structure documentation file (PHYSICSNEMO_CODE_STRUCTURE.md) to help developers and users understand the PhysicsNeMo repository organization.

What's Included

• Complete Directory Structure: Visual tree of the entire codebase
• 20+ Model Architectures: Detailed documentation of AFNO, FNO, GraphCast, Diffusion models, MeshGraphNet, Transolver, and more
• Neural Network Layers: Documentation of 50+ layer types including attention, convolution, Fourier layers, GNN layers, etc.
• Data Pipelines: Climate, CFD, Healthcare, Geophysics data loading utilities
• Metrics: Domain-specific evaluation metrics
• Distributed Computing: Multi-GPU and domain parallelism documentation
• Examples Overview: Guide to 776 example files across multiple domains
• Testing Infrastructure: Overview of 281 test files
• Getting Started Guide: Quick installation and usage instructions

Benefits

• Helps new contributors understand the codebase structure
• Provides a centralized reference for all components
• Documents the relationship between different modules
• Includes file statistics and dependency information

File Details

• Location: \PHYSICSNEMO_CODE_STRUCTURE.md\
• Size: 20.8 KB
• Format: Markdown with table of contents

[coreyjadams]

I like the idea of this, but I think it's not something we can do cleanly with AI. We also have pieces in flight that it are missing (mesh, diffusion, nn). Probably, we should have this in our documentation instead, no?