feat: table ocr confidence added

[himanshu-josh]

This PR adds confidence scoring capabilities to the table processing pipeline, enabling both cell-level and table-level confidence metrics to be captured and exposed in the JSON output. This enhancement provides valuable metadata about the quality and reliability of table extraction results.

• Added confidence: float | None = None field to TableCell class. Enables storage of individual cell confidence scores from OCR/text extraction

• Added table_confidence: float = 0.0 field to BaseTable class. Enables storage of aggregated table-level confidence scores Affects all table types: Table, TableOfContents, Form

• Added calculate_table_confidence() method. Calculates average confidence across all cells in a table

• Added calculate_cell_confidence() method. Calculates average confidence for individual table cells

• Updated assign_ocr_lines() method. Preserves confidence data from OCR results: {"text": t, "confidence": cell_text.confidence}

• Added block-specific JSON output classes:

  1. JSONTableCellOutput: Contains only confidence field
  2. JSONTableOutput: Contains only table_confidence field

• Updated extract_json() method. Implements block-specific confidence field inclusion
  TableCell blocks → JSONTableCellOutput with confidence field
  Table blocks → JSONTableOutput with table_confidence field
  Other blocks → JSONBlockOutput with no confidence fields

  This feature could be useful when someone is working with surya plus a fallback model, so based on confidence fallback to that model can be implemented.

[github-actions]

CLA Assistant Lite bot All contributors have signed the CLA ✍️ ✅

[himanshu-josh]

I have read the CLA Document and I hereby sign the CLA