added preprocessing wizard for datasets

[Thorfinson]

This PR adds an interactive data preprocessing wizard that enables users to upload, clean, and process CSV/Parquet files directly in the browser without requiring command-line tools.

Users can now prepare datasets through a guided workflow:

• Upload Data - Drag-and-drop CSV/Parquet files with automatic profiling and quality scoring
• Select Columns - Choose which columns to include with visual statistics (sparklines, box plots)
• Clean Data - Configure handling for missing values, outliers (IQR/Z-Score methods), and duplicates
• Configure Features - Set up encoding (One-Hot/Label), scaling (Standard/MinMax/Robust), and glyph mapping (exactly 5 features with smart suggestions)
• Projection Settings - Enable/configure PCA and t-SNE dimensionality reduction with customizable parameters
• Review & Process - Review configuration, execute processing, and load results into the dashboard

Key Features

• Real-time data profiling using Pyodide worker (pandas + scikit-learn in WebAssembly)
• Smart defaults based on data types (automatic encoding/scaling method selection)
• Variance-based feature suggestions for glyph visualization mapping
• Progress tracking with live updates during processing
• Configuration export/import as JSON for reproducibility
• Session persistence via localStorage
• File support: CSV and Parquet (up to 150MB)