Merge pull request #7 from HuangPuStar/feat/version-upgrade

- **Document Parsing Enhancements:**
  - Sync latest deepdoc from ragflow with improved PDF parsing capabilities
  - Fix PDF figure extraction issues and position information handling
  - Enhance LLM adapter with FenixAOS integration for better compatibility
  - Create local common modules to replace RAGFlow dependencies

- **Multi-Format Output Support:**
  - Add support for multiple output formats (txt, md, json) in parse_document_with_output
  - Implement write_text_output, write_markdown_output, and write_json_output functions
  - Add html_table_to_markdown helper with proper type checking for BeautifulSoup elements
  - Update document parser adapter description to reflect new output format capabilities

- **Cross-Page Handling:**
  - Support simple cross-page elements (single page span)
  - Revert page number modifications to maintain original DeepDoc behavior
  - Implement page number normalization in FenixAOS instead of deepdoc-lib
  - Remove unused merge_cross_page_positions function (designed for older deepdoc versions)

- **Code Quality & Infrastructure:**
  - Resolve merge conflicts and restore LLM adapter exports
  - Batch update imports to use local common and llm_adapter modules
  - Fix all import path issues and add missing modules
  - Clean up redundant files and fix import references
  - Rename test file to resolve naming conflict (test_model_resolver.py → test_model_resolver_utils.py)
  - Remove empty position_utils.py file after cleaning up unused functions

- **Current Limitations:**
  - Support simple cross-page elements but cannot handle consecutive multi-page tables/figures yet
  - Complex table structures spanning multiple non-consecutive pages require further development

- **Dependencies:**
  - Update deepdoc-lib dependency to commit dcd2dfa
  - Convert rag/res/deepdoc model files from LFS to regular file storage
  - Add rag_tokenizer backward compatibility aliases

All tests pass and mypy errorsre resolved. The system now provides robust document parsing with position information and multiple output formats.

Author: sqhyz55