feat(openai_agents): expose real `usage`, `response_id`, plumb `previous_response_id`, opt-in `prompt_cache_key` for stateful responses and prompt caching #1423

Job	Run time
find-tutorials	6s
test-00_sync/030_langgraph	4m 11s
test-00_sync/010_multiturn	2m 38s
test-10_async/00_base/010_multiturn	2m 9s
test-10_async/10_temporal/020_state_machine	2m 26s
test-00_sync/020_streaming	2m 35s
test-10_async/00_base/000_hello_acp	2m 4s
test-10_async/00_base/100_langgraph	2m 28s
test-10_async/10_temporal/010_agent_chat	3m 6s
test-10_async/10_temporal/000_hello_acp	2m 2s
test-10_async/00_base/040_other_sdks	3m 57s
test-10_async/00_base/030_tracing	2m 3s
test-10_async/10_temporal/030_custom_activities	2m 4s
test-00_sync/000_hello_acp	1m 57s
test-10_async/10_temporal/070_open_ai_agents_sdk_tools	1m 57s
test-10_async/10_temporal/050_agent_chat_guardrails	2m 3s
test-10_async/00_base/080_batch_events	2m 50s
test-10_async/10_temporal/090_claude_agents_sdk_mvp	1m 49s
test-10_async/10_temporal/060_open_ai_agents_sdk_hello_world	1m 52s
test-10_async/00_base/020_streaming	2m 8s
test-10_async/10_temporal/080_open_ai_agents_sdk_human_in_the_loop	1m 49s
Test Summary	10s
	48m 24s

Provide feedback