Reorganize project into modular architecture with separate configs and reusable components

  ## New Structure
  - configs/ - Script-specific configuration files (4 files)
    - download.py - YouTube download settings
    - build_manifest.py - Transcript processing config
    - extract_mediapipe.py - MediaPipe extraction config
    - extract_mmpose.py - MMPose 3D extraction config

  - src/asl_prep/ - Core library modules
    - common/ - Shared utilities (files.py, video.py)
    - download/ - YouTube download logic
    - transcripts/ - Text preprocessing
    - pipeline/ - Task orchestration
    - extractors/ - Landmark extraction (base, mediapipe, mmpose)

  - scripts/ - Executable entry points
    - 1_download_data.py - Download videos & transcripts
    - 2_build_manifest.py - Build manifest CSV
    - 3a_extract_mediapipe.py - Extract MediaPipe landmarks
    - 3b_extract_mmpose.py - Extract MMPose 3D landmarks

  - assets/ - Demo files and metadata (moved from resource/)

  ## Key Improvements
  - Eliminated ~200 lines of duplicate code (FPSSampler, file utils, validation)
  - Separated concerns with script-specific configs instead of monolithic conf.py
  - Created reusable modules following DRY principle
  - Added comprehensive documentation with type hints and docstrings
  - Implemented abstract base class for landmark extractors
  - Improved error handling and logging throughout

  ## Breaking Changes
  - Old scripts (s1_*.py, s2_*.py, s3_*.py) replaced with new scripts/ directory
  - conf.py split into 4 separate config files in configs/
  - resource/ directory renamed to assets/

  ## Files Changed
  - Created: 23 Python modules + 8 __init__.py files (~2,500 lines)
  - Updated: .gitignore, README.md
  - Moved: resource/ → assets/

Author: balaboom123

.gitignore

@@ -163,6 +163,20 @@ cython_debug/
 # Claude
 .claude/
 
-# data
-dataset
+# ============================================================================
+# ASL Dataset Preprocessing - Project-Specific Ignores
+# ============================================================================
+
+# Dataset directories (ignore content, keep structure)
+dataset/
+
+# Model checkpoints (large files, download separately)
+models/checkpoints/*.pth
+
+# Test notebooks and temporary scripts
+test.py
+
+# Temporary files
+*.tmp
+.DS_Store