fix(clickhouse-client): set Arrow string format options to match Cloud defaults

Jean-Gabriel Doyonrequested to merge
fix/clickhouse-arrow-format-settings into main

Summary

  • Explicitly sets output_format_arrow_string_as_string = 1 and output_format_arrow_fixed_string_as_fixed_byte_array = 1 on the ClickHouse HTTP client
  • These are the defaults on Cloud's ClickHouse 24.2+ but may differ on other environments
  • Ensures ClickHouse returns Arrow Utf8/String types instead of Binary, which is what all our extraction code (StringArray downcasts) expects

Test plan

  • Integration tests pass with settings set to 1 (42/44, 2 pre-existing flaky network failures)
  • Integration tests fail with settings set to 0 — 5 additional failures from StringArray downcast mismatches, confirming the setting is load-bearing

Merge request reports