Add read_rdf_prefixes() table function for RDF namespace introspection

[nonodename]

Summary

This PR adds a new read_rdf_prefixes() table function that extracts @prefix and @base declarations from Turtle and TriG RDF files. This enables namespace introspection, documentation generation, and building CURIE-aware tooling without parsing full RDF triples.

Key Changes

• New table function read_rdf_prefixes(): Reads one or more Turtle or TriG files and returns their prefix/base declarations as rows with columns: prefix (VARCHAR, NULL for @base), uri (VARCHAR), and is_base (BOOLEAN)

• File format support:

  • Supports Turtle (.ttl) and TriG (.trig) formats
  • Rejects NTriples and NQuads with clear error messages (these formats have no prefix declarations)
  • Auto-detects format from file extension or accepts explicit file_type parameter

• Parameters:

  • path: File path or glob pattern (required)
  • strict_parsing: Boolean flag (default true) to control error handling for malformed content
  • file_type: Override format detection (values: ttl, turtle, trig)
  • include_filenames: Boolean flag (default false) to add a 4th filename column showing source file

• Implementation details:

  • Uses libserd for efficient prefix extraction without full RDF parsing
  • Pre-computes all rows during global initialization for thread-safe scanning
  • Handles @base declarations (which have no prefix name) by emitting SQL NULL in the prefix column
  • Supports glob patterns for batch processing multiple files
  • Proper error handling with file validation at bind time

• Documentation: Added comprehensive function documentation to docs/functions.md and usage examples to README.md

• Tests: Added 20+ test cases covering basic functionality, glob patterns, file type validation, error cases, and parameter combinations

Notable Implementation Details

• Uses SERD callbacks (PrefixBaseCallback, PrefixNameCallback) to extract only prefix declarations without parsing full RDF statements
• Pre-computed PrefixRow vector approach ensures thread-safe, single-threaded scanning with atomic position tracking
• File type validation happens at bind time to provide early error feedback
• Integrates seamlessly with existing RDF extension infrastructure (reuses ITriplesBuffer::FileType and detection logic)

https://claude.ai/code/session_01VZXDXseqTYk3vvxypTwQYw