feat: add Lance-namespace integration for automatic table resolution

[beinan]

Summary

Implements automatic table resolution for Lance-backed graphs, enabling users to execute Cypher queries without manually managing table dictionaries. Addresses issue #90.

Changes

1. Convention-Based Schema Inference (LanceGraphStore.infer_graph_config())

• Automatically infers graph configuration from Lance table schemas
• Tables with single *_id field → Node tables
• Tables with 2+ *_id fields → Relationship tables

2. Store-Based Query Execution (execute_with_store())

• New function for executing Cypher queries with a LanceGraphStore
• Automatic configuration resolution (YAML schema → convention-based inference)
• Performance optimization: only loads tables required for the query
• Avoids full dataset enumeration on cloud storage

3. Comprehensive Test Suite

• 14 new tests in test_namespace.py
• Covers inference, execution, and integration scenarios
• All existing tests continue to pass (no regressions)

Usage Example

from lance_graph import CypherQuery, execute_with_store
from knowledge_graph import LanceGraphStore, KnowledgeGraphConfig

# Connect to a Lance namespace (local or cloud)
store = LanceGraphStore(KnowledgeGraphConfig.from_root("s3://my-bucket/graph-data"))

# Execute queries without manually passing table dictionaries
query = CypherQuery("MATCH (p:Person)-[:KNOWS]->(f) RETURN p.name, f.name")
result = execute_with_store(query, store)  # Automatically infers config!

Benefits

1. Minimal code changes: Builds on existing LanceGraphStore infrastructure
2. Backward compatible: Existing code continues to work unchanged
3. Performance optimized: Smart table loading prevents unnecessary data reads
4. Cloud storage ready: Works with S3/GCS via existing fsspec integration
5. Flexible: Supports YAML schema, conventions, or explicit configuration

Test Results

• ✅ 14/14 new tests passed
• ✅ 33/33 existing tests passed
• ✅ 47/47 total tests passing

🤖 Generated with Claude Code