[Storage] Implement high-performance in-memory indexed StatementStore (SPO, POS, OSP) #614

Summary

Implement a full native indexed in-memory storage in core.next.storage featuring three complete index orders (SPO, POS, OSP) to replace linear scans in InMemoryStatementStore and achieve sub-second execution on complex join queries.

Objectives and Technical Scope

1. Triple and Quad Index Structures

• Implement three primary index arrangements: Subject-Predicate-Object (SPO), Predicate-Object-Subject (POS), and Object-Subject-Predicate (OSP).
• Provide first-class support for named graphs (quads) and default graph partitioning.
• Ensure atomic lookup and range scans returning deterministic iterations.

2. Query Engine Integration

• Wire index-backed statement iterators into core.next.query solution generation pipeline.
• Eliminate linear table scan fallbacks for basic graph pattern matching.

3. Transaction Support

• Provide transactional isolation semantics (isolation of in-flight write operations, commit and rollback guarantees).

Acceptance Criteria

[ ] SPO, POS, and OSP index lookup operations are fully operational.
[ ] Named graphs (quads) and default graphs are isolated and searchable via pattern matchers.
[ ] Complex BGP join queries execute in sub-second times on datasets exceeding 50,000 statements.
[ ] Zero regressions across the W3C conformance suite.