[Documentation]: Add comprehensive Doxygen-style documentation to origami.py and matmul.py

[Copilot]

This PR adds comprehensive Doxygen-style documentation to the core TritonBLAS API files as requested in issue #X. The documentation provides complete API reference material for both users and contributors.

Files Updated

include/tritonblas/origami.py

• Module documentation: Added detailed file header explaining the hardware-aware optimization purpose and architecture
• Constants: Documented dtype_to_str mapping with inline comments for each PyTorch data type
• MatmulHeuristicResult class: Complete class documentation including:
  • Comprehensive overview of optimization algorithms and hardware considerations
  • Supported AMD GPU architectures (gfx950, gfx942/MI300X/MI300A, gfx908/MI200)
  • Matrix dimensions, data types, and execution modes explanation
• Methods: Full documentation for all public and private methods with detailed parameter descriptions, algorithm explanations, and cross-references

include/tritonblas/matmul.py

• Module documentation: Added detailed file header explaining the high-level API features and architecture
• Constants: Documented all module-level variables including device properties, buffer sizes, and pre-allocated tensors
• Functions: Complete documentation for all functions including:
  • _make_matmul_selector(): LRU caching mechanism and performance benefits
  • persistent_matmul_lt() and streamk_matmul_lt(): Detailed execution mode comparisons
  • matmul_lt() and matmul(): User-facing APIs with usage examples and code snippets

Documentation Features

• Doxygen compliance: Uses proper Doxygen syntax (@brief, @param, @return, @details, @see, @throws, @pre, @note)
• Cross-references: Extensive linking between related functions and classes for easy navigation
• Performance insights: Detailed explanations of optimization algorithms, caching benefits, and hardware-specific considerations
• Practical examples: Code snippets demonstrating proper API usage
• Hardware compatibility: Complete AMD GPU architecture support matrix with data type compatibility

Usage Example

The documentation now enables users to understand the full API:

import torch
import tritonblas

# Create matrices
A = torch.randn(1024, 512, device="cuda", dtype=torch.float16)
B = torch.randn(512, 1024, device="cuda", dtype=torch.float16) 
C = torch.zeros(1024, 1024, device="cuda", dtype=torch.float16)

# High-level API with automatic optimization
result = tritonblas.matmul(A, B, C)

# Enable Stream-K for better load balancing
result = tritonblas.matmul(A, B, C, enable_streamk=True)

This documentation will significantly improve the developer experience and reduce the learning curve for new contributors to the TritonBLAS project.

Fixes #X

Original prompt

---

This section details on the original issue you should resolve

<issue_title>[Documentation]: Need In-Code Function/API Documentation</issue_title>
<issue_description>### Description of errors

Document the code for the following files (use doxygen standard):

• https://github.com/ROCm/tritonBLAS/blob/main/include/tritonblas/origami.py
• https://github.com/ROCm/tritonBLAS/blob/main/include/tritonblas/matmul.py</issue_description>

Comments on the Issue (you are @copilot in this section)

Fixes #7

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.