feat(openai_agents): expose real `usage`, `response_id`, plumb `previous_response_id`, opt-in `prompt_cache_key` for stateful responses and prompt caching #1442

Job	Run time
find-tutorials	4s
test-00_sync/030_langgraph	2m 50s
test-00_sync/020_streaming	2m 27s
test-00_sync/000_hello_acp	2m 31s
test-10_async/00_base/010_multiturn	2m 18s
test-10_async/00_base/040_other_sdks	3m 47s
test-10_async/10_temporal/020_state_machine	3m 1s
test-00_sync/010_multiturn	3m 31s
test-10_async/10_temporal/000_hello_acp	1m 57s
test-10_async/00_base/080_batch_events	2m 55s
test-10_async/00_base/020_streaming	1m 49s
test-10_async/10_temporal/050_agent_chat_guardrails	1m 51s
test-10_async/10_temporal/090_claude_agents_sdk_mvp	1m 54s
test-10_async/10_temporal/030_custom_activities	1m 54s
test-10_async/00_base/100_langgraph	2m 9s
test-10_async/00_base/030_tracing	1m 54s
test-10_async/10_temporal/060_open_ai_agents_sdk_hello_world	1m 57s
test-10_async/10_temporal/010_agent_chat	4m 24s
test-10_async/10_temporal/070_open_ai_agents_sdk_tools	1m 57s
test-10_async/10_temporal/080_open_ai_agents_sdk_human_in_the_loop	2m 5s
test-10_async/00_base/000_hello_acp	1m 57s
Test Summary	6s
	49m 18s

Provide feedback