Load testing with new Kafka offset logic and update documentation

[AndrewAct]

What was wrong

• Kafka's offset did not work and the KafkaDeserializer did not function as expected

Fix

Replace the broken KafkaSource.builder() approach with Flink SQL metadata columns:

  CREATE TEMPORARY TABLE raw_orderbook_stream (
      payload         STRING,
      kafka_partition INT    METADATA FROM 'partition' VIRTUAL,
      kafka_offset    BIGINT METADATA FROM 'offset'    VIRTUAL
  ) WITH ('format' = 'raw', ...)

Bridged to DataStream via t_env.to_append_stream(). Real partition/offset still stored in Iceberg.

Also cleaned up:

• Deleted flink/jobs/lib/kafka_schema.py (Phase 5 merge leftover, should've been removed)
• Added flink/.dockerignore and root .dockerignore — host Python 3.13 pycache was being copied into Python
  3.10 Flink containers

Test plan

• 196 tests pass
• All 6 init containers exit 0; full stack healthy
• Airflow + Spark E2E — pending