docs: streamline 2.4.1 release notes

Release 2.4.0: Shallow index default

bump: version 2.3.2 - ugrep search improvements

Release v2.3.1

Critical Bug Fix Release

- Server shutdown AttributeError with JSONIndexManager save_index method
- Resolves crashes when using langchain or other clients that trigger server shutdown
- Remove redundant save_index call in lifespan cleanup (index auto-saves during build)

- Index persistence handled automatically in JSON-based indexing system
- No manual save_index() calls needed as index saves during build_index() operations
- Affects all MCP client integrations including langchain adapters

- No breaking changes
- Existing functionality preserved
- All indexing and search features work as expected

This is a critical fix for production deployments using MCP client libraries.

Release v2.3.0 - Major Performance Optimizations

Performance Improvements:
- Python strategy: 3 AST traversals to 1 single-pass visitor (5-10x faster)
- Java strategy: 2 tree-sitter traversals to 1 unified pass (3-5x faster)
- TypeScript strategy: 2 traversals to 1 optimized pass (3-5x faster)
- System-level parallel processing with ThreadPoolExecutor
- O(1) symbol lookup replaces O(n^2) operations

Expected Results:
- Python-heavy projects: 5-10x indexing speed improvement
- Java/TypeScript projects: 3-5x faster processing
- Large codebases (1000+ files): Dramatic reduction in indexing time
- Real-world test: 990 Java files indexed in 3-5 seconds

Technical Details:
- Single-pass AST/tree-sitter parsing across all strategies
- Hash table-based symbol lookups eliminate quadratic complexity
- Thread-safe parallel file processing
- Automatic file watcher with optimized rebuild performance

Tested and Verified:
- Comprehensive test suite with large synthetic projects
- Automatic reindexing functionality confirmed working
- Backward compatibility maintained
- All optimizations transparent to end users

This release transforms code-index-mcp into a high-performance solution suitable for enterprise-scale codebases.

Release v2.2.1: Fix file watcher auto-refresh functionality

Release v2.2.0: Return to simplified custom indexing architecture

Major architectural refinement moving away from complex SCIP protocol back to efficient custom JSON indexing.

Key Changes:
- Direct tree-sitter imports with fail-fast error handling
- Eliminated fallback mechanisms in specialized parsing strategies
- Enhanced parsing accuracy with pure AST analysis for 7 core languages
- Streamlined architecture prioritizing performance and maintainability
- Comprehensive testing across 67 sample files validates parsing improvements

This release represents a strategic return to simplicity, focusing on the project's
core mission of providing fast, accurate code indexing for LLM applications.

Enhance SCIPSymbolAnalyzer with external symbol import extraction

- Updated _organize_results to include scip_index for external symbol extraction.
- Added _extract_imports_from_external_symbols method to handle imports from SCIP index.
- Implemented framework name extraction from symbol strings.
- Classified external symbols into standard_library and third_party categories.
- Removed obsolete inspect_doc_symbols.py script.
- Updated version of code-index-mcp to 2.1.2 and added libclang dependency.
- Removed tree-sitter-c dependency from the project.

Release v2.1.1: SCIP Java Strategy Fixes

chore: prepare release v2.1.0

- Update version to 2.1.0 in pyproject.toml
- Add comprehensive release notes to CHANGELOG.md
- Build distribution packages (wheel and source)