perf(memory): explore compact typed storage and lazy cell stringification for large results (#605)

[ZhuchkaTriplesix]

Summary

Introduces columnar compact typed dataset storage (CompactResultDataset) with lazy on-demand cell stringification for large query result sets.

Changes

1. CompactResultDataset & Column Storage (lib/core/database/compact_result_dataset.dart):
   • Int64CompactColumn: Backed by Int64List and bitwise null-masking for 64-bit integer values (8 bytes per value vs 32+ bytes for full String object in heap).
   • Float64CompactColumn: Backed by Float64List and bitwise null-masking for 64-bit floating point numbers.
   • StringCompactColumn: Backed by deduplicated String references using StringInternPool.
   • GenericCompactColumn: Fallback column for composite objects and dynamic records.
2. Lazy Row List Interface:
   • asLazyRowList(): Exposes a transparent List<List<String>> view formatting cells on-the-fly when indexed by viewport-based virtual grids, preventing upfront allocation of millions of String objects.
3. Tests & Benchmarks:
   • Unit tests in test/core/database/compact_result_dataset_test.dart verifying typing, null bitmasks, lazy access, and exceptions.
   • 100,000-row x 5-column benchmark (500,000 cells) demonstrating <300ms packing and sub-millisecond visible viewport slicing.

Closes #605