Add multi-language interface support and gitignore

[jaminmc]

• Add comprehensive multi-language support (Chinese, English, Spanish, German, Japanese, French, Portuguese, Russian)
• Add language detection and localization utilities in wan/utils/language_utils.py
• Update gradio interface with dynamic language switching
• Add standard Python .gitignore file
• Replace hardcoded 'path/StableAvatar' with dynamic SCRIPT_DIR variable
• Use $(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd) to get script location
• Update all path references in inference.sh and multiple_gpu_inference.sh
• Makes scripts portable and work regardless of installation location

Optimize requirements.txt installation order for efficiency

• Reorganize packages by dependency hierarchy and installation efficiency
• Group packages logically: build tools → PyTorch → computer vision → ML frameworks → audio/video → utilities
• Install large dependencies (PyTorch ecosystem) first to avoid conflicts
• Install core numerical libraries (numpy) early as many packages depend on them
• Add clear section comments explaining the installation strategy
• Remove tokenizers (automatically installed as dependency of transformers)
• Remove imageio-ffmpeg (automatically installed as dependency of imageio[ffmpeg])
• Streamline requirements.txt by removing packages that are automatically installed
• Add documentation for optional packages that enhance performance/features

Add comprehensive optional packages section to README

• Add new 'Optional Packages for Enhanced Performance' section
• Document all optional packages: flash-attn, xformers, bitsandbytes, audio-separator[gpu], decord
• Include clear descriptions of what each package provides
• Add note about automatic fallbacks when packages are not installed
• Update existing flash_attn references for consistency
• Help users understand which optional packages to install for their use case