ematix-flow: declarative table management + load strategies
(append / truncate / merge / SCD2) for Postgres, MySQL,
SQLite, DuckDB, Delta Lake, object stores, and Kafka /
RabbitMQ / Pub/Sub / Kinesis streaming. Optional distributed
batch-SQL execution via the bundled flow-worker peer mesh.

TPC-H headline (single-host, M3 Pro):
- SF=1 full 22-query suite: DataFusion 5.87× faster than
  PySpark local[*] geomean (range 1.78× to 16.74×).
- SF=10 representative set (Q1/Q3/Q6/Q19): 3.3× geomean.

Cross-host distributed-plan numbers stay deferred (no real
cluster hardware in this project's runway). Single-node
streaming (Phase 39) is fully shipped; distributed streaming
(Σ.D) is researched + deferred until demand.

Full changelog: CHANGELOG.md [0.1.0]