GH-4428: (Enhancement) Fix typos and enhance TextSplitter with Document property preservation and chunk tracking

[Hyeri1ee]

Summary

Fix typos in spring-ai-commons and enhance TextSplitter to preserve Document properties and add chunk tracking for better RAG system support.

Current Issues

1. Typos:
   • IdGenerator.java: "dependant" should be "dependent"
   • TextSplitter.java: Comment contains "slit" instead of "split"
2. Property Loss: Document score values are lost during splitting
3. Missing Traceability: Cannot track which original document chunks came from
4. Unimplemented TODO: The "TODO copy over other properties" comment was never implemented

Changes Made

1. Typo Fixes

• IdGenerator.java: Fixed "dependant" → "dependent" in JavaDoc comment
• TextSplitter.java: Fixed "slit" → "split" in comment

2. TextSplitter Enhancements

• Preserve original document score values in all chunks
• Add tracking metadata: parent_document_id, chunk_index, total_chunks
• Implement the TODO comment for copying document properties

3. Test Updates

• TextSplitterTests.java: Added new tests and updated existing ones
• TokenTextSplitterTest.java: Updated test assertions for new metadata fields
• Test Comments: Fixed misleading comments about chunk-specific fields

Enhanced Metadata Example

//Before: Basic metadata only
{"source": "document.pdf"}

//After: Enhanced with tracking
{
  "source": "document.pdf",
  "parent_document_id": "doc-123",
  "chunk_index": 0,
  "total_chunks": 3
}

Benefits

• Document Reconstruction: Group and sort chunks by parent
• Score Preservation: Maintain relevance rankings
• Better RAG: Enhanced search context and debugging
• Code Quality: Fixed typos improve documentation accuracy

Testing

• All existing tests pass with backward compatibility
• New comprehensive test coverage added:
  • testScorePreservation(): Verifies score preservation
  • testParentDocumentTracking(): Validates parent tracking
  • testChunkMetadataInformation(): Tests chunk position metadata
  • testEnhancedMetadataWithMultipleDocuments(): Multi-document scenarios
• Updated existing tests to handle new metadata fields:
  • TextSplitterTests.java: Fixed test assertions for enhanced metadata
  • TokenTextSplitterTest.java: Updated to validate chunk-specific fields

Files Modified

• IdGenerator.java: Typo fix in JavaDoc
• TextSplitter.java: Typo fix + enhanced functionality
• TextSplitterTests.java: Updated tests + new test coverage
• TokenTextSplitterTest.java: Updated tests for enhanced metadata validationcoverage

Use Case

Document doc = Document.builder()
    .text("Long content...")
    .score(0.95)
    .build();

List<Document> chunks = splitter.split(doc);
chunks.get(0).getScore(); // → 0.95 (preserved)
chunks.get(0).getMetadata().get("parent_document_id"); // → original doc ID
chunks.get(0).getMetadata().get("chunk_index"); // → 0

This enables proper chunk-to-document relationships essential for effective RAG systems while improving overall code quality.

closes #4428