Add .clang-format for consistent C++/CUDA code style

[yurekami]

Summary

Add clang-format configuration for consistent C++/CUDA code formatting.

Configuration Highlights:

• Base: Google style with project-specific adjustments
• Indentation: 4 spaces (matches existing code)
• Line limit: 120 characters
• Braces: Attached (same line as control statement)
• Standard: C++17
• Includes: Preserve existing order (no auto-sorting)
• CUTLASS: Support for CUTLASS pragma macros

Also Includes:

• Update .gitignore to specifically ignore .clangd and .clang-tidy instead of all .clang* files

Usage

# Format a single file
clang-format -i csrc/file.hpp

# Format all C++/CUDA files
find csrc -name "*.hpp" -o -name "*.cpp" -o -name "*.cu" -o -name "*.cuh" | xargs clang-format -i

Integration

Works with pre-commit hooks (PR #252) to automatically format code on commit.

Test plan

• Configuration is valid YAML
• Style matches existing codebase
• Verify formatting output on sample files

🤖 Generated with Claude Code