feat: replace pdfjs with tabula for better pdf table extraction

[DavidAmunga]

Description

This PR replaces the PDF.js-based PDF parsing with Tabula (Java-based table extraction library) to significantly improve the accuracy and reliability of M-PESA statement parsing. This change provides better table structure recognition and handles various statement formats more consistently.

Major Changes

1. New Tabula-based PDF Processing

• Added tabulaService.ts - New service that uses Tabula for table extraction via Rust backend
  Extracts tables directly to CSV format, preserving structure better than text extraction
  Improved header detection: parser now finds the header row first and ignores all preceding content
  Better handling of quoted fields and complex CSV data

2. Removed PDF.js Dependencies

• Removed pdfService.ts (524 lines) - Old PDF.js-based parser
• Removed pdfjs-setup.ts - PDF.js worker configuration
• Removed pdfjs-dist npm dependency
• Simpler codebase with fewer dependencies

3. Enhanced Paybill Statement Support

• Added support for Paybill-specific fields:
• transactionType - Type of transaction
• otherParty - The other party involved in the transaction
• Updated CSV export to include paybill fields when present
• Updated XLSX export to dynamically add paybill columns when needed
• Automatic detection of paybill vs personal statements

Testing Recommendations

• Test with personal M-PESA statements
• Test with paybill M-PESA statements
• Test with password-protected PDFs
• Test with various statement formats and date ranges
• Verify CSV and XLSX exports include all relevant fields