fix(harness): correct codex & openai reasoning stream envelopes #1811

Job	Run time
find-tutorials	6s
test-10_async/00_base/010_multiturn	2m 12s
test-00_sync/030_langgraph	3m 2s
test-00_sync/000_hello_acp	2m 13s
test-10_async/00_base/020_streaming	2m 21s
test-00_sync/020_streaming	2m 52s
test-00_sync/040_pydantic_ai	2m 34s
test-10_async/10_temporal/050_agent_chat_guardrails	2m 6s
test-10_async/00_base/030_tracing	2m 6s
test-10_async/00_base/110_pydantic_ai	2m 16s
test-00_sync/010_multiturn	2m 46s
test-10_async/00_base/040_other_sdks	3m 51s
test-10_async/00_base/100_langgraph	2m 22s
test-10_async/10_temporal/010_agent_chat	3m 1s
test-00_sync/050_openai_agents	2m 30s
test-10_async/00_base/140_codex	2m 22s
test-10_async/10_temporal/070_open_ai_agents_sdk_tools	2m 18s
test-10_async/00_base/000_hello_acp	2m 50s
test-10_async/00_base/120_openai_agents	2m 30s
test-00_sync/060_claude_code	2m 20s
test-10_async/00_base/130_claude_code	2m 31s
test-00_sync/070_codex	2m 34s
test-10_async/10_temporal/080_open_ai_agents_sdk_human_in_the_loop	2m 17s
test-10_async/10_temporal/000_hello_acp	2m 22s
test-10_async/10_temporal/130_langgraph	2m 7s
test-10_async/10_temporal/020_state_machine	2m 25s
test-10_async/10_temporal/030_custom_activities	2m 3s
test-10_async/00_base/080_batch_events	3m 9s
test-10_async/10_temporal/120_openai_agents	2m 15s
test-10_async/10_temporal/060_open_ai_agents_sdk_hello_world	2m 18s
test-10_async/10_temporal/090_claude_agents_sdk_mvp	2m 6s
test-10_async/10_temporal/110_pydantic_ai	2m 7s
test-10_async/10_temporal/150_codex	2m 25s
test-10_async/10_temporal/140_claude_code	2m 17s
Test Summary	6s
	1h 21m 40s

Provide feedback